HUD Inspections: Steps Needed to Address Uncertainty in Inspection Scores
(Letter Report, 11/08/2000, GAO/GAO-01-109).

The Department of Housing and Urban Development (HUD) created the Real
Estate Assessment Center (REAC) to conduct physical inspections of HUD's
public and multifamily housing properties to determine whether the
properties are safe, decent, and in good repair. Although REAC uses
sampling to select units and buildings for inspection, it did not, at
the time of GAO's review, routinely calculate the sampling error
associated with the physical condition score assigned to each property.
Under REAC's sampling procedure, buildings and units do not always have
the same chance of being included in the inspection sample. REAC
officials told GAO that they were introducing new sampling procedures
that would nearly equalize the chance that each property's dwelling
units would be selected. The actions that HUD is proposing should
address that. GAO's concerns about HUD's sampling procedures.

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

 REPORTNUM:  GAO-01-109
     TITLE:  HUD Inspections: Steps Needed to Address Uncertainty in
	     Inspection Scores
      DATE:  11/08/2000
   SUBJECT:  Public housing
	     Data integrity
	     Housing programs
	     Rental housing
	     Building inspection
	     Building codes
	     Quality assurance
	     Statistical data
	     Statistical methods

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Testimony.                                               **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-01-109

A

Report to the Secretary of Housing and Urban Development

November 2000 HUD INSPECTIONS Steps Needed to Address Uncertainty in
Inspection Scores

GAO- 01- 109

Lett er

November 8, 2000 The Honorable Andrew M. Cuomo The Secretary of Housing and

Urban Development Dear Mr. Secretary: In 1997, the Department of Housing and
Urban Development (HUD) created the Real Estate Assessment Center (REAC) to,
among other things, administer physical inspections of HUD's public and
multifamily housing properties. REAC's new inspection system uses data
gathered by inspectors to calculate scores that measure the physical
condition of HUD's properties- that is, the extent to which the properties
are safe, decent, and in good repair. REAC then uses these scores, which
range from 0 to 100, to determine what follow- up is needed to address the
problems identified

during the inspections. Soon, REAC will also use these scores to determine
when properties should be reinspected. The new system does not require the
inspection of all units and buildings within each property because HUD's
housing inventory is large and its resources for inspections are limited;
instead, for most properties, the system provides for inspecting samples of
units and buildings. Sampling introduces a degree of uncertainty, called
sampling error, which statisticians commonly express as a range associated
with numerical results, such as REAC's property inspection scores. Sampling
error is a routine and accepted outcome of sampling.

We reported on various aspects of REAC's physical inspection program in July
2000. 1 Although we found that REAC's system improved on HUD's previous
inspection systems, we also found that weaknesses in the system's quality
assurance procedures limited REAC's effectiveness in ensuring the
reliability of the inspections. We also found some sampling and statistical
issues that affect the reliability of the inspection scores. As agreed with
the requesters of our July report, we are transmitting this report to you on
the impact of sampling on inspection scores. Specifically,

this report examines (1) how REAC accounts for sampling error when
calculating and reporting inspection scores and (2) whether the formula

1 HUD Housing Portfolios: HUD Has Strengthened Physical Inspections but
Needs to Resolve Concerns About Their Reliability (GAO/ RCED- 00- 168, July
25, 2000).

that REAC uses to calculate inspection scores is consistent with the
sampling procedures it uses to select buildings and units.

Results in Brief Although REAC uses sampling to select units and buildings
for inspection, it did not, at the time of our review, routinely calculate
the sampling error associated with the physical condition score assigned to
each property. REAC did, however, study the effects of sampling for about
8,800 out of 40, 000 properties. Our analysis of these data indicates that
the sampling error for about 71 percent of the properties was 3 or fewer
points, but for less than 1 percent of the properties, it was much higher-
from 8 to 19 points. Further analysis identified instances when the sampling
error was great enough to create uncertainty about whether the properties'
inspection scores fell above or below the administrative thresholds that HUD
uses to determine what follow- up actions its field offices should take or
when the property should be reinspected. For example, we found such

uncertainty associated with about 8 percent of the multifamily properties
whose inspection scores we analyzed. HUD recently agreed to take actions to
address our concerns about its treatment of sampling error. These actions
include routinely calculating the sampling error for each inspection score
and revising the procedures for determining what key follow- up actions are
needed when the sampling error would otherwise create

uncertainty about the appropriate action for the property. When implemented,
these actions should address our basic concerns.

Under the sampling procedure that REAC uses as part of its physical
inspection process, buildings and units do not always have the same chance
of being included in an inspection sample. This sampling procedure is
acceptable, but the formula REAC uses to calculate inspection scores is not
consistent with the procedure. As a result, some inspection scores are

inaccurate. According to REAC's analysis of data for about 5,000 properties,
correcting the formula to account for the probability of units' and
buildings' inclusion in a sample changed the scores of about 30 percent

of the properties inspected. For most of the properties whose scores
changed, the change was minimal; however, for about 1.4 percent of the
properties, the scores changed by 4 to 9 points- enough, potentially, to
affect REAC's decisions about the type of follow- up and the frequency of
reinspection. HUD's plans for changing its sampling and scoring processes
should, when implemented, address our concerns.

Background To ensure that families living in rental housing that is owned,
insured, or subsidized by HUD have decent, safe, and sanitary
accommodations, REAC

conducts annual inspections of public housing and multifamily properties.
For its first round (or baseline) of inspections, REAC inspected over 3,100
public housing authorities with 13, 607 properties and 26, 528 privately
owned multifamily properties- a total of 40,135 properties. Public housing
serves low- income families, the elderly, and persons with disabilities and
is operated by public housing authorities using funds provided by HUD.
Privately owned multifamily housing includes properties that receive some
form of rental assistance, including Section 8 assistance, from HUD;
properties whose mortgages are insured or held by HUD; and properties that
are financed by HUD. 2

During an inspection, an inspector enters observations into a hand- held
computer and then electronically submits the data to REAC for verification
and calculation of a score ranging from 0 to 100. Using this scoring range,
HUD establishes administrative thresholds to determine what follow- up
actions its field offices need to take and how soon the properties need to
be reinspected. The thresholds for multifamily and public housing properties
differ somewhat. Multifamily properties with scores below 60 are subject

to greater follow- up requirements than properties with scores of 60 or
higher, and properties with scores of 30 or below are referred to the
Departmental Enforcement Center. Furthermore, multifamily properties that
score below a certain threshold will be reinspected annually, whereas
properties with higher scores will be reinspected less frequently. In
November 2000, HUD expects to issue a final rule with the thresholds for
reinspections of multifamily properties.

The follow- up requirements for public housing are more complex. A public
housing authority receives a physical condition score for each of its public
housing properties. The properties' scores are then combined into an overall
indicator of the physical condition of the housing authority's properties.
The indicator for physical condition is then combined with three other
indicators, derived from assessments of the housing authority's financial
health, management operations, and residents' satisfaction, to arrive at an
overall score for the housing authority. Starting June 30, 2000,

2 Currently, HUD inspects housing that receives Section 8 project- based
assistance (“ projectbased” assistance is attached to the
structure, whereas “tenant- based” assistance is attached to the
resident).

HUD's policy has been to use this overall score to determine what followup
actions its field offices need to take, as well as how soon the housing
authority's properties should be reinspected. (Previously, HUD treated the
scores for public housing authorities as advisory- i. e., nonbinding. 3 )
When

a housing authority receives an overall score below 70, HUD will generally
require it to develop a formal improvement plan, and when an authority
receives an overall score below 60, HUD will designate it as
“troubled” and refer it to the Troubled Agency Recovery Center.
Conversely, when a housing authority receives an overall score of at least
90, it may be designated as a “high performer.” High- performing
housing authorities that also receive a physical condition score of at least
90 will be reinspected every other year, rather than annually. The purpose
of the thresholds for both multifamily and public housing is to allow HUD to
concentrate its monitoring resources on properties with lower scores.

Most physical inspection scores are based on inspections that cover a sample
of units or buildings. The results for the sample of units or buildings are
then used to estimate a score that represents the condition of the entire
property. REAC's protocol requires inspectors to inspect a statistical
sample of units or buildings. HUD officials told us that sampling was the
most cost- effective way to ensure that all properties get inspected.
However, sampling introduces some uncertainty into the final inspection
score. This uncertainty, or lack of precision, can be estimated using a
measure called sampling error. Because REAC employs a sampling procedure
(called probability sampling) to select the sample of units and buildings
used to estimate a

property's physical condition score, each property's estimated score has a
measurable precision, or sampling error, which may be expressed as a plus/
minus figure. A sampling error indicates how closely HUD can reproduce from
a sample the score that HUD would have obtained if all units and buildings
at the property had been inspected. By adding the sampling error to and
subtracting it from the estimate, HUD can develop upper and lower bounds for
each property's estimated physical condition score. This range is called a
confidence interval. Sampling errors and confidence intervals are stated at
a certain confidence level- for example, 95 percent. A confidence interval,
at the 95- percent confidence level, means

3 Prior to June 30, 2000, HUD reported the scores to housing authorities for
informational purposes, requiring only that they correct any life-
threatening health and safety violations found during inspections.

that in 95 out of 100 instances, the sampling procedures that REAC used
would produce a confidence interval containing the physical condition score
that would have been obtained if all units and buildings at the property had
been inspected.

Sampling Leads to Nearly all of the properties that HUD inspects using the
new physical Uncertainty About

inspection protocol are subject to a sampling of units, and a much smaller
percentage are subject to a sampling of buildings. However, we found that,
Follow- up

in its baseline inspections, HUD did not routinely estimate or consider the
Requirements

sampling error for each score. Using the information that REAC provided for
a subset of inspected properties, we found that the sampling error was
generally no more than 3 points, but for less than 1 percent of the
properties, it was much higher- from 8 to 19 points. As a result, sampling
caused some uncertainty about the appropriate type of follow- up required
for some properties.

Sampling Error Varies Under its new physical inspection protocol, HUD
inspects a sample of units Widely for about 95 percent of the 40, 000
properties it inspects, and for about 20 percent of the properties, it
inspects a sample of buildings. We asked REAC

for information on the sampling error of the physical condition scores it
had estimated for inspected properties. REAC provided us with the data it
had used to study the effects of its current sampling procedures on 8,813
properties- 6,291 multifamily and 2, 522 public housing properties. 4 Using
REAC's information, we computed the sampling error for these properties. We
determined that the sampling error was 3 or fewer points for about 70
percent of the 8,813 properties. However, the sampling error varied widely-
from 0 to 19 points- as shown in table 1.

4 HUD's current sampling procedures require the inspection of no more than
27 units, when dwelling units are sampled, and no more than 27 buildings
with dwelling units, when buildings are sampled. The actual number of units
and buildings sampled depends on the number of units and buildings in the
property. For example, any property with more than 1, 461 units would have
27 units sampled for inspection.

Table 1: Sampling Error for 8,813 Physical Inspection Scores

Sampling error in points

Cumulative Sampling error of score Frequency Percent percent

0 1, 884 21. 4% 21.4% 1 1, 497 17. 0% 38.4% 2 1, 543 17. 5% 55.9% 3 1, 341
15. 2% 71.1% 4 1, 130 12. 8% 83.9% 5 782 8.9% 92. 8% 6 441 5.0% 97. 8% 7 143
1.6% 99. 4% 8 28 0. 3% 99. 7% 9 11 0. 1% 99. 9% 10- 19 13 0. 1% 100. 0%

Total 8, 813 100%

Note: The sampling error, rounded to the nearest whole point, is calculated
at the 95- percent confidence level for inspections using the current
maximum sample size of 27. We multiplied the standard error, supplied by
REAC, by a t- value of 2 to obtain the sampling error. REAC's samples ranged
from 5 to 27 units. The appropriate t- values were 2.776 and 2.056 when the
numbers of units sampled were 5 and 27, respectively. Because we used a t-
value that was consistently lower than

appropriate, our estimates somewhat understate the sampling error. We could
not use the appropriate t- value because the data we obtained did not
include the number of units sampled. Source: GAO's analysis of REAC's data.
Sampling Caused

Multifamily properties are placed in one of four categories, depending on
Uncertainty About Followup their inspection scores. Properties with scores
of 60 and above are required for Some Multifamily

only to correct deficiencies found during inspection, whereas properties
Properties with lower scores have additional requirements. When a property's
physical condition score is from 31 through 59, a HUD field office imposes
follow- up requirements. When the score is 30 or below, the property is
referred to the Departmental Enforcement Center. Table 2 lays out the

different requirements for multifamily properties.

Table 2: Follow- up Requirements for Multifamily Properties Based on Their
Inspection Scores

Physical inspection score Follow- up requirement

60 and above No certification or corrective plan is required unless the
field office requests one. 46 - 59 The HUD field office requires the
property owner to inspect all

units and buildings and develop a plan of correction. The plan should give
the results of the owner's own inspection, list corrections already made to
identified deficiencies, discuss plans to correct the remaining
deficiencies, and list the resources to be used by the owner to make
repairs. 31- 45 In addition to complying with the requirements for
properties that received scores from 46 through 59, the property owner

must develop a management improvement plan and complete its goals for
correcting deficiencies. The field office can require another inspection to
confirm the completion of repairs. 30 and below The property is referred to
the Departmental Enforcement

Center (DEC) for evaluation. DEC conducts an evaluation to confirm that the
property should be assigned to it. The property owner is required to develop
a plan to correct deficiencies, and the field office is required to ensure
that all health and safety deficiencies are corrected in a timely manner.
Note: At all properties, regardless of score, deficiencies found during
inspection must be corrected. Moreover, life- threatening health and safety
deficiencies are expected to be corrected within a certain time frame,
usually within a few days of the inspection.

Using information provided by REAC on 6,291 of the over 26,000 multifamily
properties it inspected, we determined the follow- up category to which each
property would have been assigned, given its reported

physical inspection score. We then determined, after considering the
sampling error of the estimated score, whether HUD could be confident that
the property would have been placed in the same follow- up category if its
score, based on a full inspection of all of its buildings and units, had

been known. For example, if a property's score, based on a sample, was 33
and the sampling error for the estimated score was plus or minus 5, the
resulting confidence interval would range from 28 through 38. HUD therefore
could not be sure whether this property belonged in the “30 and
below” or the “31- 45” category. As shown in table 3,
sampling produced uncertainty about the appropriate requirements for 7. 6
percent of the 6,291 multifamily properties tested.

Table 3: Follow- up Referrals for 6,291 Multifamily Properties Properties
whose placement in category was questionable after sampling error was
considered

Follow- up category Number of

based on physical properties placed inspection score of

in category Number Percent

60 and above 5, 684 157 2. 8% 46- 59 384 218 56. 8% 31- 45 150 90 60. 0% 30
and below 73 16 21.9%

Total 6, 291 481 7. 6%

Note: For an additional 4 to 6 percent of the multifamily properties, we
could not tell whether the scores were reliable enough to place the
properties in the correct follow- up category. The sampling error for these
properties appeared to be reported as 0 simply because no variation in
physical condition was observed among the sampled units or buildings. In
such cases, the sampling error cannot be calculated, and we could not tell
whether the score could reliably be used to determine the appropriate
follow- up actions for the property. We assumed that appropriate actions
could be determined for these properties, but, in so doing, we may have
understated the percentage of properties whose categories may be affected by
sampling error.

Source: GAO's analysis of REAC's data.

Besides using the physical condition scores to determine the follow- up
requirements for multifamily properties, HUD plans to use the scores to
determine how soon the properties will be reinspected once it issues the

scoring thresholds for reinspection. For some properties, such decisions
will also be subject to uncertainty because of sampling error.

Sampling will also cause uncertainty for some public housing properties.
Once the public housing scores become binding, HUD will use the scores to
determine the follow- up and reinspection requirements for public housing
properties. However, the effect of sampling may be lessened for two reasons.
First, the thresholds will be based on the physical condition of all
properties within an authority. Second, the thresholds will be based on a
number of factors in addition to the physical inspection score, including
the authority's financial health and management.

REAC officials told us that REAC has established many safeguards to ensure
that the physical inspection process is accurate and fair. These include
quality assurance activities at various levels of the inspection process,
technical reviews of inspections, appeals procedures, and postinspection
reviews. They said that all of these processes can be used to help address
concerns about the accuracy and reliability of inspections. However, none of
these processes directly deal with uncertainties in inspection scores
stemming from sampling error. 5

In an August 2000 meeting, REAC's Director told us that REAC was planning,
on the basis of discussions with us about sampling issues, to calculate
confidence intervals for all inspection scores for internal use. In

addition, for properties whose scores are close to the administrative
thresholds HUD uses in determining follow- up actions (30 and 60), REAC
would request that HUD field offices consider other information on the
condition of these properties to help ensure that they are placed in the
appropriate follow- up category. Finally, for other properties whose scores
are associated with unusually large sampling errors, REAC would examine the
reasons for the large sampling errors and either take actions to reduce

the errors or alert the appropriate HUD field offices if there is
uncertainty about what follow- up actions should be taken. In its October
13, 2000, letter commenting on our draft report, HUD stated that it plans to
use an alternative approach for addressing our concerns about sampling
error. HUD is now proposing that when the confidence interval for a
property's inspection score crosses one of its key

administrative thresholds for determining follow- up actions (30 or 60
points), it will adjust the score downward so that the property will be
treated as if its score falls into the follow- up category requiring more
intensive monitoring. While this approach differs somewhat from the approach
that HUD proposed in the August 2000 meeting and that we had discussed with
REAC officials, we believe that it should still address our basic concerns.
6

5 Our July report discusses other concerns about the reliability of REAC's
inspection scores. 6 The approaches that we had discussed generally focused
on having HUD program offices perform additional analyses for properties
whose scores could not be relied on to ensure that appropriate follow- up
actions were taken. For example, if there were uncertainty due to sampling
error about whether a property's score was above or below 60, staff might
take into account the owner's past responsiveness in addressing physical
defects when determining whether a written corrective action plan should be
required.

We recognize that HUD's proposal focuses on two key administrative
thresholds (30 and 60 points) and does not specifically cover other
administrative thresholds that HUD uses to determine follow- up or
reinspection requirements. However, the fact that HUD will routinely
calculate the sampling error for all inspection scores should help it
determine the extent to which sampling may cause uncertainty about decisions
in these areas as well. This should put HUD in a good position to continue
identifying strategies and opportunities to mitigate the effects of

uncertainty caused by sampling. REAC's Scoring

REAC attempted to reduce the sampling error associated with its Formula Is
Not

inspection scores by giving buildings with a large number of units a greater
chance of being selected for inspection than other buildings. Although the
Consistent With Its use of such a sampling procedure is acceptable, the
formula that REAC

Sampling Technique used to calculate the inspection scores was not
consistent with this sampling approach. REAC used a formula that was
appropriate only when all units and all buildings with dwelling units at a
property had an equal

chance of being inspected. As a result, HUD incorrectly calculated some
inspection scores. HUD tested a method for calculating the score that
corrected this problem, using data for 5, 030 properties, and found that the
revision changed the scores of about 30 percent of these properties (see
table 4).

Table 4: Change in Physical Condition Score After Accounting for the
Probability of Selection for Inspection

Change in points

Percentage of Change in score Number of properties properties

0 3, 561 70. 8% 1 1, 058 21. 0% 2 258 5.1% 3 84 1. 7% 4-9 69 1. 4%

Total 5, 030 100.0%

Source: GAO's analysis of REAC's tables.

While the scores of 97 percent of the properties tested changed by no more
than 2 points, the scores for 69 (or 1.4%) of the 5,030 properties changed
by 4 to 9 points. As noted, when a score changes by even one point, the
followup requirements can change. Among the 69 properties whose scores
changed by 4 or more points, REAC identified 6 properties whose follow- up
requirements could have been affected. 7 For example, one multifamily
property in this analysis received a score of 64. Given this score, HUD
would not have required a corrective plan for the property. However, when
the differing probabilities of buildings' and units' selection for
inspection were considered, the property received a score of 58 points. With
this score, the HUD field office would require the owner to inspect all of
the

property's units and buildings and develop a corrective plan to ensure the
repair of all deficiencies. REAC agrees that, for complete accuracy, the
scores estimated from samples of units and buildings should account for
differing probabilities of selection. In an August meeting, REAC officials
told us they were introducing new sampling procedures that would nearly
equalize the

chance that each of a property's dwelling units would be selected for
inspection. They said that information obtained during the baseline
inspections made the new sampling method feasible.

In its October 13, 2000, letter commenting on our draft report, HUD agreed
to take further actions to address our concerns. HUD said it would change
its sampling procedures to obtain a self- weighting sample of units and

modify the inspection- scoring algorithm to include the probability of
selection for sampled buildings. These actions, when implemented, should
address our concerns in this area. Conclusions We understand REAC's need to
sample buildings and units when performing physical inspections and
recognize the inherent uncertainty associated with sampling. However, it is
important for REAC to recognize

both the extent to which this uncertainty affects its inspection scores and
the need for appropriate procedures to mitigate the effects of the
uncertainty. The actions that HUD is proposing to take in response to our

concerns about its sampling procedures should, we believe, basically address
these concerns. 7 REAC did not provide information on how frequently the
changes in the scores for the other 1, 400 properties affected their follow-
up requirements.

Agency Comments and We provided HUD with a draft of this report for its
review and comment.

Our Evaluation HUD said that to resolve our concerns, it has revised its
procedures to ensure that all inspected properties are properly and
consistently

monitored in accordance with risk management practices and it will alter
scoring- related processes to increase precision. More specifically, to
address the recommendations we made in our draft report, HUD stated that it
would begin calculating the confidence interval for every inspection score.
In addition, HUD said that when the confidence interval for a property's
inspection score crossed one of the key administrative thresholds for
determining follow- up actions (30 or 60 points), it would adjust the score
downward so that the property would be

treated as if its score fell into the follow- up category requiring more
intensive monitoring by HUD. According to HUD, this action would reduce or
eliminate the risk of inadequate monitoring. HUD noted that it expected this
action would affect only about 3 percent of its multifamily properties and
that the benefits of increased scrutiny would exceed the inconvenience to a
small number of property owners.

HUD also agreed to change its data collection, scoring, and reporting
processes to incorporate new sampling procedures in its inspection data
collection device to obtain a self- weighting sample of units, modify the

inspection- scoring algorithm to include the probability of selection for
sampled buildings, and modify its inspection summary report to show the
probability of selection for sampled buildings. We revised our report to
include the actions HUD is proposing. We also dropped our draft report's
recommendations for improvements in these areas, since HUD's actions, when
implemented, should address our basic concerns about HUD's sampling
procedures. The complete text of HUD's comments appears in appendix I.

Scope and To determine whether REAC's procedures were producing inspection
Methodology

scores that could reliably be used to determine the appropriate follow- up
requirements for inspected properties, we reviewed REAC's current sampling
procedures and discussed these procedures with the appropriate REAC
officials. We also obtained REAC's calculations of standard errors and
scores for 8,813 properties- including 6,291 multifamily and 2,522 public
housing properties- inspected using the current sampling procedures. We
calculated the sampling error for each property using its

inspection score and standard error. From the 6,291 multifamily properties,
we identified those for which the sampling error introduced uncertainty
about the appropriate follow- up action.

To determine whether the method used to compute scores was correctly
estimating the score a property would have received if the entire property,
and not just a sample of its dwelling units or buildings, had been
inspected, we reviewed REAC's documentation for sampling procedures and
formula for calculating scores. We obtained REAC's analysis of how the
scores for 5,030 properties would change if the formula were changed. We
also discussed REAC's procedures and analysis with appropriate REAC
officials. We conducted our work from December 1999 through October 2000 in
accordance with generally accepted government auditing

standards. We are sending copies of this report to Honorable Rick Lazio,
Chairman, and the Honorable Barney Frank, Ranking Minority Member,
Subcommittee on Housing and Community Opportunity, House Committee on
Banking and Financial Services, who requested our July 2000 report on HUD's
physical inspection system. We will also make copies of this report
available to others on request. If you or your staff have any questions
about this report, please call me at (202) 512- 7631. Key contributors to
this report

were Karen Bracey, Martha Chow, and Richard Hale. Sincerely yours,

Stanley J. Czerwinski Director, Physical Infrastructure Issues

Appendi xes Comments From the Department of Housing

Appendi x I and Urban Development

Lett er (385874) Lett er

Ordering Information The first copy of each GAO report is free. Additional
copies of reports are $2 each. A check or money order should be made out to

the Superintendent of Documents. VISA and MasterCard credit cards are
accepted, also. Orders for 100 or more copies to be mailed to a single
address are discounted 25 percent.

Orders by mail: U. S. General Accounting Office P. O. Box 37050 Washington,
DC 20013

Orders by visiting: Room 1100 700 4th St. NW (corner of 4th and G Sts. NW)
U. S. General Accounting Office Washington, DC

Orders by phone: (202) 512- 6000 fax: (202) 512- 6061 TDD (202) 512- 2537

Each day, GAO issues a list of newly available reports and testimony. To
receive facsimile copies of the daily list or any list from the past 30
days, please call (202) 512- 6000 using a touchtone phone. A recorded menu
will provide information on how to obtain these lists.

Orders by Internet: For information on how to access GAO reports on the
Internet, send an e- mail message with “info” in the body to:
info@ www. gao. gov or visit GAO's World Wide Web home page at: http:// www.
gao. gov

To Report Fraud,

Contact one:

Waste, or Abuse in Web site: http:// www. gao. gov/ fraudnet/ fraudnet. htm

Federal Programs

e- mail: fraudnet@ gao. gov 1- 800- 424- 5454 (automated answering system)

GAO United States General Accounting Office

Page 1 GAO- 01- 109 HUD Inspections United States General Accounting Office

Washington, D. C. 20548 Page 1 GAO- 01- 109 HUD Inspections

Page 2 GAO- 01- 109 HUD Inspections

Page 3 GAO- 01- 109 HUD Inspections

Page 4 GAO- 01- 109 HUD Inspections

Page 5 GAO- 01- 109 HUD Inspections

Page 6 GAO- 01- 109 HUD Inspections

Page 7 GAO- 01- 109 HUD Inspections

Page 8 GAO- 01- 109 HUD Inspections

Page 9 GAO- 01- 109 HUD Inspections

Page 10 GAO- 01- 109 HUD Inspections

Page 11 GAO- 01- 109 HUD Inspections

Page 12 GAO- 01- 109 HUD Inspections

Page 13 GAO- 01- 109 HUD Inspections

Page 14 GAO- 01- 109 HUD Inspections

Appendix I

Appendix I Comments From the Department of Housing and Urban Development

Page 15 GAO- 01- 109 HUD Inspections

Appendix I Comments From the Department of Housing and Urban Development

Page 16 GAO- 01- 109 HUD Inspections

United States General Accounting Office Washington, D. C. 20548- 0001

Official Business Penalty for Private Use $300

Address Correction Requested Bulk Rate

Postage & Fees Paid GAO Permit No. GI00
*** End of document. ***