United States
Environmental
Protection Agency
Office of
Pollution Prevention
and Toxics
EPA747-R95-003
April 1995
Report on the National
Survey of Lead-Based
Paint in Housing
Base Report
-------
REPORT ON THE NATIONAL SURVEY
OF LEAD-BASED PAINT IN HOUSING
Base Report
This work was conducted under
HUD Contract Number HC-5848 and
EPA Contract Numbers 68-D9-0174,68-D2-0139 and 68-D3-0011
June, 1995
-------
Hie material in this document has been subject to Agency technical and policy review and approved
for publication as an EPA report. The views expressed by individual authors, however, are their
own and do not necessarily reflect hose of the U.S. Environmental Protection Agency. Mention of
trade names, products, or services does not convey, and should not be interpreted as conveying,
official EPA approval, endorsement, or recommendation.
-------
ACKNOWLEDGMENTS
Westat would like to acknowledge the many people that contributed to this study. The data
analysis and report writing work was directed by Manie Chen, Work Assignment Manager, Office of
Pollution Prevention and Toxics (OPPT), U.S. Environmental Protection Agency (EPA), Phil Robinson,
Section Head, Technical Programs Branch, and Cindy Stroup, Chief; Technical Programs Branch, OPPTS,
EPA.
Westat appreciates the support and guidance of staff from EPA, the U.S. Department of
Housing and Urban Development (HUD), and Cox Associates far reviewing the research and providing
technical guidance and input In addition to the above, they include: John Schwemberger, Bradley Schultz,
and Joseph Breen of EPA; Stevenson Weitz, Ellis Goldman, Ron Morony and Art Newburg of HUD; and
Arnold Greenland and David Cox of Cox Associates.
Westat carried out the research for this study and wrote this report Key members of the
Westat research team include: Robert Clickner, Project Director and Co-Principal Investigator; John
Rogers, Co-Principal Investigator, Victoria Albright, Field Director, Richard Rinehart, Research
Associate; Charles Carusi, Research Associate; Edward Bryant, Statistician; Janice Machado, Research
Associate; William Devlin, Research Associate; Stephen Dietz, Corporate Officer, John Michael,
Statistician; Ben Topping, Statistician; Valerija Smith, Statistician; George Fisher, Programmer; David
Lowe, Programmer; Helen Powell, Programmer; Joyce Powell, Programmer; Angela HoSang, Research
Assistant; Mary Madigan, Research Associate; Debbie Khchell, Editor; Gudrun Tomasson, Research
Assistant; Michael Vendetti, Research Assistant; and Brian Dingwall, Research Assistant
The national survey of lead-based paint in housing was conducted for HUD by Westat, Lie.,
in affiliation with Dewberry and Davis and KTA-Tator, under HUD Contract HC-S848. Initial data
analyses were conducted by Westat, Inc. in affiliation with Speedwell, Inc., under the same contract The
additional analyses reported in this report were conducted by Westat under EPA Contracts 68-D9-0174,
68-D2-0139 and 68-D3-0011.
-------
TABLE OF CONTENTS
ACKNOWLEDGMENTS i
EXECUTIVE SUMMARY v
1 INTRODUCTION 1-1
1.1 Background 1-1
1.2 Report Organization 1-3
1.3 Objectives of National Survey of Lead-Based Paint in Housing 1-3
1.4 Lead Standards and Guidelines 1-4
2 NATIONAL ESTIMATES OF PREVALENCE 2-1
2.1 Major Findings and Discussion 2-1
2.2 Lead Paint Hazard in Private Housing 2-2
2.3 Lead Paint Hazard in Public Housing 2-9
3 RESEARCH DESIGN 3-1
3.1 Sample Design. 3-1
3.2 Inspection Protocol 3-4
3.3 Environmental Sampling Protocols 3-5
3.4 Quality Assurance 3-6
3.4.1 Measures to Enhance Response Rates 3-6
3.4.2 Quality Assurance for MAP/XRF Data 3-6
3.4.3 Field Quality Assurance Measures 3-7
3.5 Limitations of the Study Design 3-7
4 SOURCES OF ERROR IN THE NATIONAL SURVEY DATA 4-1
4.1 Response Rates and Potential for Non-Response Bias 4-1
4.1.1 Private Housing 4-1
4.1.2 Public Housing 4-2
4.2 MAP/XRF Measurement Bias 4-3
4.3 Classification Error Due to MAP/XRF Measurement Variation 4-3
4.4 Classification Error Due to Dwelling Unit Sampling Procedures 4-4
4.5 Impact on the National Estimates of Prevalence 4-4
11
-------
TABLE OF CONTENTS (continued)
Chapter
5 CONCLUSIONS 5-1
5.1 Major Conclusions 5-1
5.1.1 Study Findings 5-1
5.1.2 Impact nf Measurement F.rrnr and Lead JfeuKng Variation
on the Data Analysis 5-1
5.1.3 Use of the Spectrum Analyzer MAPyXRF 5-2
5.2 Additional Conclusions and Recommendations 5-2
List of Tables
Table Page
2-1 Estimated Number of Privately Owned Occupied Housing Units
Built Before 1980 with Lead-Based Paint, by Selected Characteristics
(Paint Lead Concentration-1.0 mg/«q cm) 2-3
2-2 Estimated Number and Percent of Public Housing Units Built Before
1980 with Lead-Based Paint, by Selected Characteristics
(Paint Lead Concentration^ 1.0 mgkq cm) 2-10
3-1 National Distribution of Dwelling Units Built Before 1980 3-2
3-2 Distribution of Completed Inspections by Construction Year and
Dwelling Unit Type 3-2
4-1 National Response, Contact, and Eligibility Rates at Each Data
Collection Stage 4-2
List of Figures
Figure Page
2-1 Prevalence of Lead-Based Paint by Location in Privately Owned
Occupied Housing Units 2-4
2-2 Estimated Percent of Privately Owned Housing Units with Dust
and Soil Lead Exceeding Guidelines by Construction Year 2-5
2-3 Number of Privately Owned Housing Units with Interior Dust
Lead Levels Above Guidelines by Presence and Condition of
Lead-Based Paint (LBP) 2-6
2-4 Number of Privately Owned Housing Units with Lead in Soil
by Presence and Condition of Exterior Lead-Based Paint 2-7
2-5 Average Amount of Lead-Based Paint (LBP) Per Housing Unit
on Interior Surfaces by Painted Component for Privately
Owned Housing Units (LBP Concentration >- l.Omg/igcm) 2-8
in
-------
Figure
TABLE OF CONTENTS (continued)
List of Figures (continued)
2-6 Average Amount of Lead-Based Paint (LBP) Per Privately Owned
Housing Unit on Exterior Surfaces by Painted Component
(LBP Concentration > = l.Omg/sgcm) 2-8
2-7 Amounts of Lead-Based Paint by Location and Construction Year
fiDr Privately Owned Housing Units (LBP coocentniion>=i.o mg/tg em) 2-11
2-8 Geometric Mean Paint Lead Loadings in Privately Owned Housing
Units by Location and Construction Year (LBP coocemiuioo>-1.0 ag/tg cm) 2-11
2-9 Percent of Public Housing Units with Lead-Based Paint (LBP)
by Location (LBP CoooodntiaD>-1.0 mg/ig cm) 2-12
2-10 Average Amounts of Lead-Based Paint (LBP) per Public Housing
Unit on Interior Surfaces by Architectural Component
(LBP Concentration >-1.0 mgfcg on) 2-13
2-11 Average Amount of Lead-Based Paint (LBP) in Publicly Owned
Housing Units on Exterior Surfaces by Architectural Component
(LBP Concentration > = 1.0 mg/tg cm) 2-13
2-12 Geometric Mean Paint Lead Loadings in Publicly Owned Housing
Units by Location and Construction Year (LBPConcentration> - i.omg/sgcm) 2-14
IV
-------
EXECUTIVE SUMMARY
Lead is a powerful toxin that attack? the central nervous system and is particularly Imaging to the
developing nervous systems of young children. High levels of lead in the blood can result in convulsions,
mental retardation, and even death. Further, recent medical research has found that low levels of lead
exposure have more serious health consequences that previously thought. Effects include reductions in
intelligence and short-term memory, slower reaction times, and poorer hand-eye coordination.1
Although there are many sources of lead in the environment., including drinking water, food,
emissions from gasoline combustion, and industrial emissions, it is clear that lead-based paint plays a
major role in high blood lead levels. Recent research indicates that dust and soil may be the most
significant pathway for low-level lead exposure, and that lead-based paint is an important source of
household dust lead."
The 1987 amendments to the Lead-Based Paint Poisoning Prevention Act required the Secretary of
Housing and Urban Development (HUD) to prepare and transmit to Congress "a comprehensive and
workable plan" for the abatement of lead-based paint in hni|g'ng and "an estimate of the amount,
characteristics and regional distribution of housing in the United States that contains lead-based paint
hazards at differing levels of contamination." In response to this mandate HUD sponsored a national
survey of lead-based paint in housing and delivered a Report to Congress on a Comprehensive and
Workable Plan for the Abatement of Lead-Based Paint in Privately Owned Housing in December, 1990.
The Comprehensive and Workable Plan report was completed under a tight, Congressionally mandate
schedule and focused on motivating, developing and presenting the comprehensive plan required by
Congress. As such, it only reported the estimates of the extent of lead-based paint in housing required by
Congress and provided a brief description of the survey methodology.
This report, sponsored by the Environmental Protection Agency, is a comprehensive technical report
on the HUD-sponsored national survey of lead-based paint in housing. It provides a detailed description of
the survey methodology. It reports on wide ranging analyses of the national survey data. It reports revised
estimates of the extent of lead-based paint in housing, based on a thorough investigation of the multiple
sources of error - variability and bias - in the data. These error sources include nonresponse biases,
sampling variability between housing units, sampling variability within housing units, X-ray fluorescence
device (XRF) measurement error, and laboratory analysis error. The analysis underlying the estimates
presented in the Comprehensive and Workable Plan (CWP) report incorporated only sampling variability
between housing units.
Extent of Lead-based Paint in Housing - Revised Estimates
The number of housing units classified as having lead-based paint depends on the definition
employed to classify a housing unit as having lead-based paint The definition used here classifies a home
Fowler, Brace et al, "Measuring Lead Exposure in Infants, Children, and Other Sensitive Population!,* National Academy Press,
Washington, D.C.
... Hammond P. B., Dietrich K. N., Succop P., Krafft K.. Clark S., Berger O., Peanon D., and Que Hee S. (1985). The
Cincinnati prospective study of low-level lead exposure and its effects on child development: Protocol and sttnis report. Environmental
Research 38,4-18.
Bornschein R.L., Succop P. A., Dietrich K. N., dark C. S., Hee Q. S. and Hammond P. B. (1985). The influence of social and
environmental factors on dust lead, hand lead, and blood lead levels in young children. Environmental Research 38, 108-118.
-------
as having lead-based paint if the measured lead concentration on any painted surface is 1.0 mg/cm2, or
greater.
As reported in the CWP report, lead-based paint is widespread in housing. The revised estimate is
that 64 million homes (± 7 million)4, 83 percent (± 9%) of the privately owned housing units built before
1980, have lead-based paint somewhere in the building. (Fifty-seven million (± 5 million) homes, or 74
percent (± 6%), were reported in the CWP.) Twelve million (± 1 million) of these homes are occupied by
families with children under the age of seven years old. An estimated 49 million (± 7 million) privately
owned homes have lead-based paint in their interiors. There are no statistically significant differences in
the prevalence of lead-based paint by type of housing, market value of the home, amount of rent payment,
household income, or geographic region.
Seventeen percent of the pre-1980 housing stock have dust lead levels in excess of the federal
guidelines5, independent of the presence or absence of lead-based paint. However, excessive dust lead
levels are associated with the presence of damaged lead-based paint. Fourteen million homes, 19 percent
of the pre-1980 housing stock, have more than five square feet of damaged lead-based paint. Nearly half of
them (47 percent) have excessive dust lead levels.
Excessive soil lead levels6 are also associated with the presence of damaged lead-based paint. While
21 percent of all pre-1980 homes have excessive soil lead levels, nearly half of the 10 million homes with
non-intact lead-based paint on exterior walls have excessive soil lead levels.
Although a large majority of pre-1980 homes have lead-based paint, most of them have relatively
small areas of it The average privately-owned housing unit with lead-based paint has an estimated 601
square feet of it on interior surfaces and 869 square feet on exterior surfaces. Over half of the leaded paint
is on walls, ceilings, and floors. (For comparison, the walls in a room 10' by 12', with an 8' ceiling, have
an area of 352 square feet) The amounts of lead-based paint per housing unit vary with the age of the
dwelling unit Pre-1940 units have, on average, about three times as much lead-based paint as units built
between 1960 and 1979.
Lead paint is even more widespread in public housing; 86 percent (± 8%) of all pre-1980 public
housing family units have lead-based paint somewhere in the building. While most public housing units
have some lead-based paint, most of them have small areas of surfaces covered with it. The average public
housing unit with lead-based paint has an estimated 367 square feet on interior surfaces and 133 square
feet on exterior surfaces. Most of the interior lead-based paint is on walls, while very little of the exterior
walls are painted.
Survey Methodology
The objective of the national survey of lead-based paint in housing was to obtain data for estimating:
(1) the number of housing units with lead-based paint; (2) the surface area of lead-based paint in housing,
to develop an estimate of national abatement costs; (3) the condition of the paint; (4) the prevalence of lead
in dust in housing units and in soil around the perimeter of residential structures; and (5) the characteristics
of housing with varying levels of potential hazard, to examine possible priorities for abatement
The number* in parentheses are 95% confidence intervals.
The federal guideline* vary with the location of the dust: 200 nucrograms of lead per aquare foot for floor samples, 500 Mg/tq ft for
window aill* and 800 |ig/sq ft for window well*.
5
EPA interim guideline* for coil lead action level* are 500 ppm for residential toil associated with Super&nd sites.
VI
-------
The study population consisted of nearly all housing in the United States constructed before 1980.
Vacant housing, group quarters, Alaska and Hawaii were excluded for operational reasons. Newer houses
were presumed to be lead-free because, in 1978, the Consumer Product Safety Commission banned the sale
of lead-based paint to consumers and the use of such paint in residences. Hie survey was conducted
between December 1989 and March 1990 in 30 counties across the 48 contiguous states, selected to
represent the entire United States housing stock, both public and privately-owned. The total sample size is
381 dwelling units, 284 privately owned and 97 publicly owned. The sample was small, but it provided
estimates that were sufficiently precise to develop the CWP for private and public housing.
Within each housing unit, two rooms were randomly selected for inspection; one room with
plumbing (wet) and one room without plumbing (dry). la each of these two rooms, the field technicians
inventoried painted surfaces, measured their dimensions, and assessed the condition of the paint; they
measured the lead concentration in randomly selected painted surfaces; and they gathered samples of dust.
Since not all rooms in a dwelling unit were inspected, it is possible to miss lead-based paint when it
is really present somewhere else in the dwelling unit To reduce the chances of misclassifying a dwelling
unit with lead-based paint as lead-free, additional lead readings, termed purposive readings, were taken on
surfaces that, in the opinion of the field technicians, were most likely to have lead-based paint. In some
dwelling units, these additional purposive samples did, indeed, find lead-based paint in dwelling units where
no lead-based paint had been found in the randomly selected rooms.
Exterior painted surfaces were inventoried and measured and lead readings taken according to
protocols similar to those used in the interior. Soil samples were also taken at selected locations around the
building exterior. Common areas were also sampled and inspected.
Lead in paint measurements were made with portable Schec MAP-3 spectrum analyzer XRF devices
(MAP/XRFs), which MIST had determined to be more accurate and more precise than the direct-reading
XRFs used in earlier surveys. Although the MAP/XRFs were an improvement over the earlier direct-
reading XRFs, they still bad limitations. In particular, MAP/XRF measurements made over brick or
concrete were less accurate and less precise than those made over wood or plaster. These limitations
notwithstanding, portable MAP/XRF technology was used because the survey included occupied dwellings
where it was not feasible to take paint scrapings for laboratory analysis.
Data were collected on lead in dust and soil in a number of locations in each sampled dwelling unit
Dust samples were collected by vacuuming randomly selected floor locations, window sills and window
wells in the wet room and again in the dry room. In addition, a dust sample was collected from the floor
just inside the main entrance to the dwelling unit. Soil samples were taken outside the main entrance to the
building, at a selected location along the drip line of the sampled exterior painted surface, and at a remote
location away from the building but still on the property. Dust and soil samples were analyzed for lead
concentration.
Analyses of the Sources of Error
Nonresponse analysis. An analysis was conducted of the private housing nonresponse in the
national survey to estimate the potential for nonresponse bias. This analysis was necessary because
intrusive studies that impose significant burdens on the respondents tend to have lower response rates than
less burdensome studies and, therefore greater potential for nonresponse biases. In this survey, S3 percent
of the homes asked to permit the inspection visits cooperated fully with the study. The nonresponse
analyses did not reveal any evidence of potential non-response biases associated with ethnicity, building
age, or family income. There were statistically significant associations between the response rates and
monthly rent in tenant-occupied housing units and current market value in owner-occupied housing; the
lower ends of both distributions were somewhat under represented. There was a strong positive correlation
vu
-------
between inspected housing units in the same census block with respect to the presence of lead in paint, dust,
and soil. On balance, these findings suggest that the potential bias due to nonresponse is likely to be small.
Correction for measurement bias. The MAP/XRF measurement equipment used to detect and
quantify lead in paint tended to yield readings that were biased, Le., systematically different from the actual
lead concentrations measured. Quality assurance (QA) data collected daily during the national survey field
period permitted the estimation of the MAP/XRF bias. MAP/XRF readings were made on shims of known
lead concentration placed over selected substrate materials. A shim is a piece of hard paper painted with
lead-based paint. There were four substrate materials, wood, drywall, steel, and concrete, selected to
represent the typical range of substrate materials encountered in residential construction. Statistical
techniques were applied to the QA data to develop calibration equations for adjusting the MAP/XRF
readings for measurement bias. The MAP/XRF TgaHingg taken in the housing units were therefore adjusted
to statistically correct for measurement bias.
Correction for misclassification errors. There are two major factors that induce misclassification
errors. First, the MAP/XRF equipment also has random variability in its measurements. This variation
ran induce a classification bias, that is, a bias in the estimated prevalence of housing units with lead-based
paint Second, the protocol for inspecting a housing unit for lead-based paint provided for sampling
painted surfaces for MAP/XRF measurement, rather than measuring the lead content of every painted
surface in tile housing unit Under this inspection protocol, it is possible for a housing unit to have some
surfaces with lead-based paint, other painted surfaces without lead-based paint, and only the lead-free
surfaces selected for MAP/XRF measurement Such housing units would be incorrectly classified as not
having lead-based paint.
To adjust for these classification biases, the distribution of lead concentrations on the untested
painted surfaces, in the sampled rooms ?nd in the unsampled rooms, in each household was simulated. (No
applicable, adequate data set existed to permit the direct estimation of these biases.) This extension from
the measured surfaces to all surfaces in the unit was based on (1) data on the number of rooms in the unit,
(2) data on the number of surfaces per room, and (3) assumptions about the relationship of the lead
concentrations on unmeasured surfaces to those on the sampled and measured surfaces. The model was
observed to be consistent with the National Survey data. The misclassification rates were estimated for the
simulated housing units, and used to adjust the prevalence estimates accordingly.
Impact on the national survey findings. All findings on the lead hazard in homes reported
elsewhere in this report incorporate the results of this error analyses. That is, the raw MAP/XRF readings
have been statistically corrected for measurement bias and the misclassifications due to measurement
variation and sampling within dwelling units have been corrected. The national estimate of prevalence of
lead-based paint in privately-owned housing is 83 percent. Without the statistical corrections described
above, the estimate would have been 74 percent.
-------
1. INTRODUCTION
The National Survey of Lead-Based Paint in Housing was conducted under the sponsorship of the
Department of Housing and Urban Development to provide basic data for comprehensive and workable
plans (CWPs) for the prompt and cost-effective inspection and abatement of lead-based paint hazards in
private and public housing. The CWPs were required by the Lead-Based Paint Poisoning Prevention Act
(LPPPA), as amended by Section 566 of the Housing and Community Development Act of 1987. ' The
CWP for private housing was issued to Congress in December, 1990 2 The CWP for public housing is
1.1 Background
This section presents the background for the national survey. It briefly describes the data available
prior to the national survey on lead-based paint hazards.
Previous Surveys
There were four notable surveys of lead-based paint in housing before the national survey. Three
local surveys, in Washington, D.C., Pittsburgh, and Phoenix, were conducted in the mid-1970s, and one
national survey of public housing was carried out in the 1980s.
The Washington, DC, survey, conducted in 1973 by the National Bureau of Standards (now the
National Institute of Standards and Technology [NIST]) had a sample of 233 housing units representing the
caty of Washington.3 This survey also acted as a HUD-sponsored field test for a Pittsburgh survey
conducted a year later.
The Pittsburgh survey, conducted in 1974 and 1975 by the Allegheny County (PA) Health
Department for the National Bureau of Standards under HUD sponsorship, is by fer the largest study of its
type ever conducted. The survey completed inspections in approximately 3,300 housing units from the
entire Pittsburgh urban area.4
The Phoenix survey, conducted in 1976 by the Arizona Department of Health Services, had a
sample of 268 units. The sample represented a single Phoenix Census tract that was chosen because of a
high number of bom pre-1940 units and children under five years old.5
Amendment in Section 566 of the Housing tod CoomjunftyDevelopeDait Act of 1987 (Public Law 100-242X
TJ.S. Department of Housing and Urban Devtkpmem, Office of Policy J>e>ndofmeAt^J(aatA(1990\Comprehenji^ imdWoricabU Plan for
Hall, William; *nd Aym, Tyrone (1974X Survey Plans and Data ColUcton and Anafyiu Mahodologits: Jtuuta efa Prt-Survey for the
Maptitudt andExuntofthfLtad-BattdPatnt Hazard in Housing (NBSIR 74-426), US. Department of Commerce. National Bureau of
Standard*.
Parts I and H (NBSIR 77-1250 and 77-1293). U.S. Department of Commerce, National Bureau of Standard*.
Arizooa Oeptrtmea cfHetJth Service*. Dnnoo of Envi^^
the State Legislature" (mourn).
1-1
-------
The fourth survey was part of a national survey conducted in 1984-1985 to address modernization
needs of public housing. Two hundred and sixty-two public housing units (apartments) plus associated
common areas (Le., hallways, playgrounds) were inspected in 131 public housing projects in 34 cities.6
Limitations of Previous Surveys
As a basis far national estimates of the number of housing units with lead-based paint, analyzing
other lead hazards in housing, and estimating the cost of abatement, these prior surveys are limited.
Sample Limitation. Because of the limited geographic coverage of most of the surveys, there is no
way of knowing the extent to which the findings are representative of housing in the nation.
Wide Divergence in Estimates of Homes with Lead. The four prior surveys' estimates for the
percentage of homes with lead varied immensely. For housing built prior to 1940, the estimates ranged
from 71 to 100 percent; for homes built between 1940 and 1959, the range was 64 to 92 percent; and for
units built between 1960 and 1977, the range was 48 to 76 percent.
Measurement Imprecision. The portable X-ray fluorescence (XRF) analyzers used to measure the
lead content of paint in all of the surveys have subsequently been found by NIST to be highly imprecise at
the 1.0 mg/cm2 (milligrams of lead per square centimeter of painted surfaces) level7, which is the Federal
threshold for lead-based paint.
Lack of Dust and Soil Lead Data. The prior surveys provide no information on the prevalence of
lead in house dust and exterior soiltwo sources identified in the literature as important pathways of lead.
Therefore, they cannot be used to analyze the prevalence of lead in dust and soil, or the association between
lead-based paint and lead in dust and soil.
Studies of Lead in Surface Dust and Soil
A large number of studies published during the past two decades have indicated an association
between lead dust and childhood blood lead. Three studies established the apparent importance of lead dust
as a pathway for lead-based paint. In 1980, Chamey and colleagues concluded that, although several
factors accounted for childhood lead poisoning, lead dust and "hand" lead (i.e., lead that clings to fingers
and hands) were strongly correlated with blood lead, and that interior lead dust should be taken into
account in attempting to reduce lead hazards in residential environments.*
In 1983, based on a HUD-funded study in Baltimore, Chamey and colleagues analyzed whether dust
control measures, in addition to treatment of potential lead-based paint hazards, would lower blood levels.
They concluded "that a focused dust-control program can reduce blood lead levels more than standard lead
removal in the home."9
Wallace. James E, (1986), The Cost of Lead-Based Paint Abatement in Public Housing, US. Department of Housing and Wan Development.
7McKnignt, Mary E.; Byrd, Erie W.; Roberts, Willed E.; and Lagergren, Eric S. (December 1989X Methods for Measuring Lead Concentrations In
Paint Films (NISTIR 89-4209), U.S. Department of Cumulate, National Institute of Standards and Technology.
8Chamey. E.; Smyre, J.; and Coulter, M. (Febnimiy 1980), "Increased Lead Absotption m fan» City Oiildro
Pediatrics, 6S<2).
9Cbaroey, E^ Kessler, B; Farfei, M.; and Jackson, D. (1983X "Childhood Lead Poisoning: A Controlled Trial of the Effect of DustControI Measure*
on Blood Lead Levels," New England Journal of Medicine 309(18):1089-1093.
1-2
-------
Bellinger and colleagues (1986) enrolled 249 metropolitan Boston children with low-to-moderate
blood lead levels at one month of age and collected a wide range of data semiannually. Environmental lead
and mouthing behavior were significantly associated with blood lead, but home environment/care giving,
child development, and sociodemographic characteristics were not. Dust lead was the most important
environmental variable, although refinishing and month of sample selection were significant.
Studies of Pathways Between Paint Lead and Blood Lead
Bomschein, et al. (1986)11 studied the relationship between children's blood lead levels and measures
of the extent of lead-based paint in dwelling units. They found that lead in paint does not directly impact
blood lead levels, but it does impact them through the pathways:
Lead-based paint hazard > dust lead > blood lead, and
Lead-based paint hazard > dust lead > hanj lead > blood lead.
The conclusion is that, except for children with pica, an abnormal craving to eat non-food
substances, dust is the major immediate source of lead for children, and that lead-based paint is a primary
contributor to dust lead.
1.2 Report Organization
In the sections that follow, this document presents the design, methodology, findings and conclusions
of the survey, and discusses quality assurance and sources of error in the survey data. The presentation
begins with discussions of the national estimates of prevalence, amounts, and hazards of lead in paint, dust,
and soil in private and public housing. Next, a discussion of research design addresses sample design, data
collection protocols, and quality assurance. Finally, this report addresses the major conclusions to be
drawn from the national survey and its findings
1.3 Objectives of National Survey of Lead-Based Paint in Housing
The objectives of the national survey were based on Section 302 of the Lead-Based Paint Poison
Prevention Act of 1971, as amended. The informational requirements set forth in that Act are:
An estimate of the amount, characteristics, and regional distribution of housing in the United
States fnat contains lead-based paint hazards at differing levels of contamination.
A comprehensive and workable plan for the cost-effective inspection and abatement of public
housing..., including an estimate of the total cost of abatement.
-------
A comprehensive and workable plan, including any recommendations for changes in
legislation, for the prompt and cost-effective inspection and abatement of privately-owned
single family and multi-family housing, including housing assisted under Section 8 of the
U.S. Housing Act of 1937.
Toward these objectives, the National Survey of Lead-Based Paint in Housing was designed to
obtain data for estimating:
The number of dwelling units in the United States with interior and exterior lead-based paint,
by year built, type of housing, level of lead loading, and census region (see Section 1.4 for an
explanation of lead loading).
The number of multi-family (private and public) residences with lead-based paint in common
areas, by year built, level of lead loading, and census region.
The extent of surface area of lead-based paint in order to estimate national abatement costs in
public and privately owned housing.
The prevalence of damaged lead-based paint.
The prevalence of lead in dust in dwelling units and in soil around the perimeter of residential
structures.
The characteristics of housing with varying levels of hazard to examine possible priorities for
abatement.
The information was needed to support a number of research questions. These included: analysis of
the relationship among sources and pathways of lead in the residential environment; analysis of the
characteristics of housing with varying hazard levels; development of indices of lead hazard; analysis of the
costs, effectiveness and benefits of alternative strategies of reducing lead-based paint hazards: and the
identification of the dimensions of each of these issues.
1.4 Lead Standards and Guidelines
The Federal Government has set standards and guidelines for lead hazards in paint, dust and soil.
They are briefly discussed here.
The amount of lead in the paint, expressed as the paint lead loading on a surface, is an important
factor in determining the potential exposure level of hazard. For this study, the Federal standard (set by
HUD) is 1.0 milligrams of lead per square centimeter of painted surface (1.0 rng/cm2), and is used to
classify a painted surface as "with" or "without" lead-based paint. Therefore, painted surfaces with 0.9
mg/cm2 of lead are classified as having no detectable lead; and painted surfaces measured at 1.0 mg/cm2
are considered to have lead. At the same time, painted surfaces with lead as high as or in excess of 5.0
mg/cm2 (more common in older homes) are classified in the same wayas having lead-based paint It
should also be noted that the Federal standard for paint lead loading is not a hearth-based standard,
implying 1.0 mg/cm2 is not necessarily a threshold for determining safe or unsafe conditions. EPA, along
with other Federal agencies, is currently developing a health-based standard for paint lead, along with
standardized measurement techniques.
Dust lead levels presented in this report are a "loading" and soil lead levels are a "concentration".
Dust loadings refer to the mass of lead per square area of surface, expressed as micrpgrams of lead per
1-4
-------
square foot of surface (ug/ft2). The HUD Guidelines state that no more than 200 ug/ft2 on floors, 500
ug/ft2 on windowsills and 800 fig/ft2 on window wells is allowed inside a home after lead paint abatement
takes place. (Windowsills are defined by HUD as the lower part of a window inside the room, and window
wells are the bottom of a window between the glass and screen.)
For dust, the guidelines are set by HUD, and were developed for clearance purposes only, i.e.,
declaring an abated residence ready for re-occupancy after lead paint abatement They were not developed
for determining the presence or absence of health hazards. la addition, readers should be cautious when
comparing the national survey dust iead results to the HUD guidelines because dust samples in the survey
were collected with a vacuum sampling technique and the HUD guidelines were developed for a wipe
sampling technique. The comparability between the two techniques is unknown, i.e., each technique may
give very different results when samples are collected side-by-side. There is evidence mat wipe samples
tend to result in higher lead loading men vacuum samples.
Soil concentration is reported as the mass of lead per unit mass of soil, usually expressed as
micrograms of lead per gram of soil (ug/g), or the equivalent parts per million (ppm). The guidelines for
soil lead used in this report (SOO ppm) were derived from the EPA's Interim Guidance on Establishing Soil
Lead Cleanup Levels at Superjund Sites and do not necessarily reflect the potential exposures to occupants
near residential structures.
Although the paint, dust, and soil guidelines are not health-based, they do provide a framework to
compare the relative differences among differing characteristics between homes, such as whether lead-
based paint is intact or non-intact. As noted for paint lead standards, EPA is currently developing health-
based lead standards for dust and soil.
1-5
-------
2. NATIONAL ESTIMATES OF PREVALENCE
This chapter presorts findings on the lead hazard in homes. Some of the estimates are revised from
the estimates reported in the Comprehensive and Workable Plan for the Abatement of Lead-Based Paint
in Privately Owned Housing: Report to Congress (CWP Report) in light of the measurement error
analyses reported in Appendix n, Chapter 3.
2.1 Major Findings and Discussion
An estimated 64 million (77 million to 81 million)13 or 83 percent (± 9%) of private housing units in
the United States built before 1980 have lead-based paint on either interior painted surfaces, exterior
painted surfaces, or both, (57 million (± 5 million), or 74 percent (± 6%) was reported in the CWP Report).
Approximately 782,000 (± 77,000) or 86 percent (± 8%) of pre-1980 public housing units also have lead-
based paint on their surfaces. Although these numbers are vast, they do not necessarily suggest that each
home is an immediate hazard to its occupants. There are many potential factors which determine the
hazards posed by lead-based paint. Several are discussed below.
In addition to the lead loading levels, the condition of the paint is also a factor in determining the
hazards from lead-based paint Except during renovations, intact lead-based paint probably poses little
immpriiatft risks to occupants; however, peeling, chipping, or otherwise deteriorating paint (non-intact) may
present an immediate danger to occupants. An estimated 14 million or 19 percent of pre-1980 private
housing units have non-intact lead-based paint on their surfaces. This is significant not only because
peeling and chipping paint may be directly ingested by children, but it is more likely to break-off and
contaminate house dust and soil. Young children ingest large amounts of dust and soil every day through
normal hand-to-mouth contact. Because of this normal activity, dust and soil are considered to be the most
significant routes of lead exposure to children. Therefore, understanding lead pathways resulting in
exposure is essential to preventing childhood lead poisoning, and was a major goal of this study.
The national survey also examined lead in house dust and in soil next to residential structures.
Thirteen million or 17 percent of private housing units have interior dust lead levels above Federal
guidelines. Similarly, 16 million, or 21 percent, of private housing units have soil lead levels above
guidelines.
National public housing estimates for non-intact lead-based paint were not calculated from the
national survey data because less than 10 homes in the survey had non-intact paint. Dust and soil lead data
also are not presented for public housing because of situations encountered during the field operations; the
dust lead data is suspect because a large number of vacant apartments were inspected. About 70 percent of
the units sampled did not have exposed soil nearby to collect samples they were surrounded by pavement
With sample sizes this small, it was not possible to project national estimates.
The following sections summarize the national survey findings. Section 2.2 discusses private
housing and Section 2.3 covers public housing. For a more complete and technical treatment of the results,
readers are referred to Appendix n, Chapter 2 of this report.
The number* in ptrcoueic* «re 95% confidence intervals. See Appendix H, Section 3.43 for the methodology wed to compute with the
2-1
-------
2.2 Lead Paint Hazard in Private Housing
Table 2-1 summarizes data collected and analyzed for the national survey. Among several
characteristics, the data suggests that older homes are more likely to have lead-based paint than newer
homes. An estimated 76 percent (± 12%) of the housing units built between 1960 and 1979 have lead-
contaminated paint on their surfaces, but the percentage increases to 90 percent (± 8%) for homes built
before 1960. Also, 12 million (± 1 million) of the homes determined to have lead-based paint are occupied
by families with children under the age of seven. This is an important statistic because childhood lead
poisoning is thought to be the most common and preventable public health concern in our country today.14
Figure 2-1 shows the prevalence of lead-based paint by location in private housing. In the figure,
the entire circle (pie chart) represents all privately-owned houses in the United States built before 1980
(about 77 million from Table 2-1). An estimated 17 percent of these have no lead-based paint at all (the
white "none" area in the pie chart), while the remainder, or 83 percent of the total, have lead-based paint
somewhere on the building. As is evident from the figure, most houses with lead-based paint have it on
both the interior and the exterior surfaces (53 percent).
It is evident from Figure 2-2 that older homes have considerably more dust and soil lead than newer
homes. The percentage of homes over the guidelines increases from less than 10 percent for post-1960
homes to about 50 percent for pre-1940 homes. House dust characteristics are further described in Figure
2-3. Loadings below and above HUD's guidelines are presented by a bar graph in relation to the location
of lead-based paint and whether it is intact or non-intact. It is important to note that the graph compares
two pieces of information: the estimated number of houses containing lead-based paint relative to the total
number of housing units projected, and the comparison between houses with dust lead within and exceeding
guidelines (the clear bar vs. the black bar). As the bar graph shows, units with intact lead-based paint are
less likely to exceed dust lead guidelines than units with non-intact lead-based paint. In fact, more than
half of the homes with non-intact exterior lead-based paint are estimated to have dust lead loadings above
HUD's Guidelines.
Figure 2-4 shows the same type of information for soil as Figure 2-3 presented for dust. Again, one
half of the private housing units in the United States with non-intact exterior lead-based paint have soil lead
concentrations above 500 ppm.
Figures 2-5 and 2-6 are pie charts that show the amounts of lead-based paint on interior and exterior
private housing surfaces for different types of painted architectural components. An estimated 29 billion
square feet or 12 percent of all painted interior surfaces are covered with lead-based paint and an estimated
49 billion square feet or 44 percent of all painted exterior surfaces are covered with lead-based paint. On
average, each private housing unit with lead-based paint has approximately 601 square feet of the lead
paint on the interior and 869 square feet on the exterior. As is evident from the figures, most of the interior
paint is on the walls, ceilings, or floors.
The amounts of lead-based paint in private housing units are further displayed in Figure 2-7. The
bar graph shows that older homes are covered with more square feet of lead-based paint than newer homes.
Figure 2-8 is similar to Figure 2-7, however, it shows the average amount of lead per square centimeter of
painted surface (more precisely, geometric mean paint lead loadings [mg/cm2]) on houses with lead-based
paint. The bar graph clearly displays older homes with more lead contamination in their paint.
I4CDC [1991]. Preventing Lead Poisoning in Young Children. U.S. Department of Health and Human Services, Public Health Service,
Center for Diieaie Control.
2-2
-------
TABLE 2-1
ESTIMATED NUMBER OF PRIVATELY OWNED OCCUPIED HOUSING UNITS
BUILT BEFORE 1980 WITH LEAD-BASED PAINT, BY SELECTED CHARACTERISTICS
(Paint Lead Concentration > = 1.0 mg/sq on)
Characteristic
Total Occupied Housing
Units Built Before 1980
Construction Yean
1960-1979
1940-1959
Before 1940
Housing Type
Single Family
Millrifumily
One or More Children
Under Age 7
Census Region
Northeast
Midwest
South
Wett
Owner-Occupied (2)
Market Value of Home
Less than $40,000
$40,000 to $79,999
$80,000 to $149,999
$150,000 and up
Renter-Occupied (2)
Monthly Rent Payment
Lew than $400
$400 and up
Household Income (2)
Less than $30,000
$30.000 and up
Total
Occupied Housing
Units (000) (1)
77,177
100%
35,681
46%
20,476
27%
21,018
27%
66,418
86%
10,759
14%
13,912
18%
16,963
22%
19.848
26%
24,967
32%
15,399
20%
50,554
11,885
24%
19,401
38%
11,863
23%
7,405
15%
24,734
16,339
66%
8,395
34%
46,126
60%
31.048
40%
Housing Units
With Lead-Based Paint
Somewhere in Building
Percent
83%
(9%)
76%
(12%)
92%
(8%)
88%
(9%)
85%
(9%)
77%
(17%)
89%
(9%)
86%
(13%)
91%
(10%)
82%
(10%)
73%
(18%)
84%
(9%)
92%
(12%)
90%
(10%)
68%
(18%)
85%
(15%)
82%
(11%)
85%
(14%)
81%
(15%)
85%
(10%)
81%
(11%)
Number (000)
64,443
(6,946)
27,275
(4,282)
18.742
(1,638)
18,424
(1,892)
56,130
(5,978)
8,308
(1.829)
12,425
(1.252)
14,605
(2.205)
18.115
(1,985)
20,393
(2.497)
11,298
(2.772)
42,516
(4.550)
10,888
(1.426)
17,550
(1.976)
8,093
(2.135)
6,276
(1.111)
20,329
(2.721)
13,811
(2.287)
6,822
(1.259)
39,032
(4,613)
25,121
(3.415)
Number of
Housing Units
in Sample (2)
284
120
87
77
227
57
90
53
69
116
46
179
39
46
45
42
105
59
40
156
127
(1) Total units data are from the 1987 American Housing Survey.
(2) Some respondents did not respond to the questions on economic variables. Therefore, counts for
disaggregation may not add to corresponding aggregate counts.
Note: Numbers in parentheses are approximate half-widths of 95% confidence intervals for the
percent! and numbers. For example, the approximate 95% confidence interval for the percent of bousing
units with some lead-based paint is 83% + / -9% or 74% to 92%.
2-3
-------
Figure 2-1
Prevalence of Lead-Based Paint by Location in Privately Owned Occupied
Housing Units
Interior Only
10%
Exterior Only
20%
Both Interior and
Exterior
53%
See Appendix n, Table A-l for underlying data
2-4
-------
Figure 2-2
Estimated Percent of Privately Owned Housing ,Units with Dust and Soil Lead Exceeding Guidelines by
Construction Year
s
Before
1940
Construction Year
(1) Surface dust lead guidelines are 200 ug/sq. ft. on floors,
500 ug/sq. ft. on window sills and 800 ug/sq. ft. in window wells
(2) Soil lead guideline levels are 500 ppm. t
(3) See Appendix II, Table 2-25, and Table 2-27 for underlying data
Soil Lead Exceeding Guidelines
Dust Lead Exceeding Guidelines
-------
Figure 2-3
Number of Privately Owned Housing Units with Interior Dust Lead Levels Above Guidelines by Presence and
Condition of Lead-Based Paint (LBP)
X
/
/
71
A
jKfi
w>
'fflfflffifa
mP~
x~
71
xf
i
n,
w<
I
I
^
i
f
/ - 7L_.
/~
71
V
^
I
*p
^s
I
i
f
/
1
«N
f
X
D Dust within guWdines
u& Dust exceeding guldeunei
No LBP
Interior
Intact
LBP
Interior Exterior
Non Intact
Intact LBP
Condition of Paint
(1) "Within Guidelines" means that surface dust lead does not exceed 200 ug/sq. ft. on floors, 500
ug/sq. ft. on window sills, and 800 ug/sq. ft. in window wells. See HUD Interim Guidelines
(2) A housing unit has non-intact interior LBP if there are more than 5 sq. ft. of damaged interior LBP.
(3) See Appendix II, Table 2-6 for underlying data
Exterior
Non
Intact
-------
Figure 2-4
Number of Privately Owned Housing Units with Lead in Soil by Presence and Condition of Exterior Lead-
Based Paint
D Soil Lead Within Guidelines
^ Soil Lead Exceeding Guidlines
No Exterior LBP Exterior Intact LBP Exterior Non-Intact
Presence and Conditon of Exterior Lead-Based Paint
(1) "Within Guidelines" means soil lead is less than SOOppm
(2) A housing unit has non intact exterior lead based paint if there are more than S sq. ft. of damaged exterior LBP.
(3) See Appendix II, Table 2-7 for underlying data
-------
Figure 2-5
Average Amount of Lead-Based Paint (LBP) Per Homing Unit on Interior Surfaces by
Painted Component for Privately Owned Housing Units
(LBP) Concentration > =. 1.0 mg/sq cm)
WaDs/CeOing/Floor
371 sq. ft.
Shelves/Other
82 sq. ft.
Non-Metal
Components
146 sq. ft.
See Appendix n, Table 2-9 for underlying data
Metal Components 2
sq. ft.
Total of 601 square feet per housing unit, average
Figure 2-6
Average Amount of Lead-Based Paint (LBP) Per Privately Owned Housing Unit on
Exterior Surfaces by Painted Component
(LBP concentrations > = 1.0 mg/sq cm)
Walls, 681 sq. ft.
Porches/ Other, 13 sq,
ft.
; Appendix n, Table 2-10 for underlying data
Non Metal
Components
169 sq. ft
Metal Components
7 sq.ft.
Total of 870 square feet per housing unit, average
2-8
-------
2.3 Lead Paint Hazard in Public Housing
Table 2-2 shows that an estimated 782,000 units, or 86 percent of the pre-1980 housing stock have
lead-based paint somewhere in the unit As with private housing, older public housing units are more likely
than newer ones to have lead-based paint. In feet, nearly all (97 percent) of pre-1950 public housing has
some lead-based paint.
Figure 2-9 shows the distribution of lead-based paint on public housing units. The data shows mat
interiors have more lead-based paint than exteriors. This is the opposite of the private housing results,
where more lead-based paint was estimated to be on the exteriors of dwellings.
While most of the interior lead-based paint in public housing (like private housing) is on the walls,
ceilings or floors (Figure 2-10), exteriors in public housing show a different trend. Figure 2-11 shows that
most of the paint on public housing exteriors is on non-metal components, including such items as trim,
window molding, doors, soffit, and fascia, (not exterior walls). This finding may reflect a majority of
public housing units with unpamted brick or concrete exterior walls.
Figure 2-12 for public housing parallels Figures 2-8 for private housing. Figure 2-12 shows that
age contributes to the amount of lead on the painted surfaces, similar to Private housing. Geometric mean
paint lead loading levels are higher for the pre-1950 dwelling units.
2-9
-------
TABLE 2-2
ESTIMATED NUMBER AND PERCENT OF PUBLIC HOUSING UNITS
BUILT BEFORE 1980 WITH LEAD-BASED PAINT, BY SELECTED CHARACTERISTICS
(Paint Lead Concentration > - l.t nog/aq on)
Characteristic
Total Pubic Homing Units Built
Before 198*
Construction Yean 1960-1979
1950-1959
Before 1950
Total
Public Housing
Units (MO)
910
100%
455
50%
273
30%
182
20%
Homing
WithLead-Bi
Ptrccfit
86%
( 78% - 94% )
79%
(66% - 92% )
90%
( 77% - 100% )
97%
( 88% - 100% )
Units
tscd Punt
D Building
Number (vM)
782
( 705 - 858 )
359
(299 - 419 )
246
(209 - 273 )
177
( 160 - 182 )
Number of
Housing Units
in Sample
97
43
24
30
Notes: (1) Numbers in parentheses are 95 % confidence intervals for the estimated percents and numbers.
(2) Categories with small sample sizes should be interpreted with caution.
2-10
-------
See Appendix n, Table A-6 & A-7 for underlying data
Figure 2-7
Amounts of Lead-Based Paint by Location and Construction Year for Privately Owned Housing
Units
(LBP Concentration > = 1.0)
CoMlructioD Year
I960 -1979
Figure 2-3
ric Mean Paint Lead iMtKnge in Privately Owned Housing Units by Location and
Construction Year
(LBP Concentration > = 1.0 mg/ sq.on)
Before 1940
1940-1959
Construction Year
1960-1979
See Appendix II, Table A-8 for underlying datt
2-11
-------
Figure 2-9
Percent of Public Housing Units with Lead-Based Paint (LBP) by Location
(LBP Concentration > = 1.0 mg/ sq. cm)
90%-r
* 2
a
'B
20%-
10%-
0%
Unit
Interior
Interior
Common
Areas
Building
Exteriors
Location
Somewhere
in
Building
See Appendix II, Table 2-30 for underlying data
2-12
-------
Figure 2-10
Average Amounts of Lead-Based Paint (LBP) per Public Housing Unit on Interior
Surfaces by Architectural Component
(LBP Concentration >= 1.0 mg/sq cm) '
Non-Metal Components
47 sq.ft.
Shelves/Other 6sq.ft.
Metal Components
32 sq. ft.
Wall/CelHng/Floor 282sq. ft
Total of 367 square feet per unit, average, interior
See Appendix II, Table 2-31 for underlying data
Figure 2-11
Average Amount of Lead-Based Paint (LBP) in Publicly Owned Housing Units on
Exterior Surfaces by Architectural Component
(LBP concentrations > = 1.0 mg/sq. cm)
Porches/Other, 4 sq. ft
Walk, 13 sq.ft.
Non-Metal Component
71 sq. ft
Metal Component
45 sq.ft.
See Appendix O, Table 2-32 for underlying data
Total of 133 square feet per unit, average, exterior
2-13
-------
, ' Figure 2-12
Geometric Mean Paint Lead Loadings in Publicly Owned Housing Units by
Location and Construction Year
(LBP Concentration > = 1.0 rag/ sq.cm)
Mean Loadings in
mg/ sq. cm
f --.,
Before 1950
1950 -1959
&'
Construction Year
1960-1979
See Appendix n, Table B-4 for underlying data
2-14
-------
3. RESEARCH DESIGN
This chapter presents a brief description of the research design for the national survey of lead-based
paint in housing. A complete technical discussion can be found in Appendix I of this report. The study
population consisted of nearly all housing in the United States constructed before 1980. Newer houses
were presumed to be no detectable lead because, in 1978, the Consumer Product Safety Commission
banned the sale of lead-based paint to consumers and the use of such paint in residences. Group quarters
(e.g., dormitories and jails) and projects that are occupied exclusively by the elderly were excluded from the
survey, because the primary public health concern was lead poisoning in children. Also excluded for
operational reasons were vacation homes, homes in Alaska and Hawaii, and military housing.
The broad elements of the design included:
A national area probability multi-stage sample survey of lead-based paint in homes, to permit
statistically valid estimates of the parameters required for the development of federal policy.
In-person visits to the sampled homes to identify, describe, and measure painted surfaces and
to measure the lead content of the paint.
Collection and laboratory analysis of dust and soil samples in and around each sampled
home, to permit the statistical analysis of the associations among lead in paint, dust and soil.
3.1 Sample Design
This section presents the sample design of the national survey. Private and public housing were
sampled, using somewhat different methodologies. This section covers both.
In order to optimize the Congressionally required estimates (see Chapter 1), a design stratified on
dwelling unit age and type was constructed. Dwelling units were grouped into three types of housing
privately-owned single family, multifamily, and public housing and three age categories. Table 3-1
displays the national distribution of pre-1980 dwelling units. The survey was conducted between
December 1989 and March 1990. The final sample contained 284 privately-owned dwelling units and 97
public housing units, for a total sample size of 381 dwelling units. Table 3-2 displays the distribution of
the final sample across the strata.
The sample design was the result of compromise between funding availability and the research
objectives. The sample was affordable and provided the statistical accuracy needed to develop national
estimates upon which government policy could be based. It provided estimates that were sufficiently
precise for the comprehensive and workable plans for private and public housing. For example, 95 percent
confidence intervals for estimates of the incidence of lead-based paint in private housing in any of the three
periods of construction used in Table 3-1 were within plus or minus 10 percent.
The multi-stage design outlined below was chosen over other approaches because it efficiently and
economically satisfied the major operational requirements of the study. The objectives and design of the
study required in-person visits to the sampled dwelling units. The geographic clustering of homes allowed
those visits to be conducted much more economically. Also, a survey sample requires a sampling frame,
i.e., a list of all dwelling units eligible for the survey. For private housing, no such list exists nationally, or
even in many localities. In light of time and cost constraints, the required number of lists was
3-1
-------
Table 3-1
National Distribution of Dwelling Units Built Before 1980
Number of Pre-1980 Dweffine Units (000)
Type
Privately Owned- Occupied
Single Family
Mnlrifnmily
Sub-Total
All Public, Family Units
Total
Construction Year
1960-1979
29,137
6448
35,685
1960-1979
182
35,867
1940-1959
18,782
1.690
20,472
1950-1959
278
20,750
pre-1940
18,499
2421
21,020
pre-1950
346
21,366
Total
66,418
10.759
77,177
Total
807
77,984
Source: 1987 American Housing Survey.
Table 3-2
Distribution of Completed Inspections
By Construction Year and Dwelling Unit Type
Completed Inspection Visits
Type
Pnrately Owned- Occupied
Single Family
MnHifamilj
Sub-Total
All Pubfic, Family Units
Total
Construction Year
1960-1979
94
26
120
1960-1979
30
150
1940-1959
72
15
87
1950-1959
24
111
pre-1940
61
16
77
pre-1950
43
120
Total
227
57
284
Total
97
381
3-2
-------
by using the multi-stage area probability sample for the private housing. A geographic clustering of the
sampled homes was deemed to be operationally efficient <*nH statistically acceptable. Also, tying the
development of the public housing list frame to the same localities as used far the private housing extended
die efficiency of the clustering of the whole survey operation.
The first stage of the sample selection consisted of selecting 30 counties stratified by region and
climate, and selected with probability proportional to the 1980 population from approximately 3,000
counties in the United States. A survey sample requires a sampling frame, that is, a list of all dwelling
units eligible for the survey. Although no such list exists for private housing, HUD maintains a list of
public housing projects suitable for use as a frame. Thus, private and public housing were sampled using
different within-county designs. The different ownership structures between public anrf private housing
also required different approaches for the owners and occupants. These two approaches are summarized
below.
Privately Owned Housing. A multistage area probability design was developed and employed to
sample private housing and approach the occupants. A description of the design and its within-county
implementation follows.
Five census blocks were statistically selected in each of the 30 counties. To ensure that the
full spectrum of income levels would be represented in the sample, a measure of wealth was
computed for each census block. One census block was selected from each one-fifth of the
range of this wealth measure.
Field interviewers were sent to each of the ISO census blocks to list every dwelling unit This
process created a frame for the sampling of dwelling units. The interviewers listed 27,833
dwelling units, an average of 186 per census block.
Samples of dwelling units were selected from the lists. A brief in-person screening interview
was conducted with an adult occupant in each sampled dwelling unit to determine if the unit
was eligible for the survey and, if so, which of the six age/type strata in Table 3-1 it belonged
to.
A sample of dwelling units was randomly selected from the eligible homes according to the
survey design in a manner that ensured that all eligible segments were represented in the
sample.
Telephone calls were made to the sampled homes to collect basic information about the
dwelling unit and the occupants and to schedule an appointment to visit the dwelling unit. An
incentive of $50 was paid to each respondent who permitted a completed inspection.
Public Housing. The public housing sample was designed in a somewhat different manner than the
private housing sample. All projects in the 30 sampled counties were extracted from the public housing
data tape supplied by HUD. The cognizant public housing authorities (PHAs) were contacted and asked to
verify and/or correct the frame data (primarily address and number of family units) for their projects. A
substantial number of additions, deletions and modifications were made to the 30-counry public housing
frame. Projects were then sampled with probability proportional to the number of family units, and
according to the construction year strata in Table 3-2. Finally, individual housing units within projects
were selected with equal probability. It was necessary to work through the chain-of-command to contact
the sampled projects and make appointments to inspect sampled units. The chain of command typically
included the HUD regional office, the PHA headquarters, and the project manager. Public housing tenants
who consented to die inspection received the same $50 incentive as private housing occupants.
3-3
-------
3.2 Inspection Protocol
Resource limitations did not permit the complete identification, inspection, quantification, and
testing of every painted surface in every inspected dwelling unit Fortunately, the objectives of the study
did not require such thorough inspections. It is possible in survey statistics to develop a good, clear picture
of the aggregate population with only limited information on each sampled individual. Consequently, the
limited inspection protocol described below was followed in each dwelling unit The inspection protocol
was die same for both public and private housing units.
The inspection visits were performed by a two-person team: a field interviewer who interviewed the
occupant and collected and recorded the information; and a technician who performed x-ray fluorescence
(XRF) testing of the lead content of the paint and collected the dust and soil samples.
Interior Rooms. The rooms were inventoried and classified into wet and dry rooms according to
the presence or absence of plumbing in the room. One wet room and one dry room were randomly selected.
All painted surfaces in each of these two rooms were identified and quantified; the substrate materials were
identified; the condition of the paint and substrate materials were noted. Painted surfaces were grouped
into four strata of architectural components: walls, ceiling, and floors; trim on metal substrates (molding,
window frames, door fiames, etc.); trim on nonmetal substrates; and other (built-in shelves, cabinets, etc.)
Five surfaces in each room were randomly selected to be tested for lead content One surface was selected
from each of the four strata, with a fifth randomly selected from all four strata. The paint lead
measurements were made with Schec MAP-3 spectrum analyzer X-ray fluorescence devices (MAP/XRFs).
Samples of dust were collected by vacuuming in three locations in each sampled room: the floor, a
window sill, and a window well. In addition, a dust sample was collected from the floor near the most-used
entrance to the building. The floor is a collection sink for dust and children come into direct contact with it
when crawling or playing. The windows are boundaries between the interior of the unit and exterior
sources of environmental lead. Because the abrasion of opening and closing windows causes paint dust to
collect on horizontal window components, die wells are collection sinks for dust. Entry ways and common
heavy traffic areas are more likely to collect deposits of soil tracked in from outside the building. The dust
samples were sent directly to a laboratory to be analyzed for their lead content.
Exterior. An exterior wall was randomly selected for inspection. All painted surfaces on the
sampled wall were cataloged in the same manner as the interior rooms and subjected to XRF testing. Three
exterior soil samples were collected: at the drip line along the sampled wall, at a remote location away
from the building, and at the most-used entrance to the dwelling unit. The entry way is a proximate source
of exterior soil lead that could be tracked into the interior. The drip line of a wall is the collection point for
run-off from painted exterior wall; flaking paint or paint components leached by rainwater can collect in
the soil. The remote location provided data on the background contribution of environmental sources of
lead not associated with the unit or building itself. The soil samples were also sent to a laboratory for
lead content analysis.
Because random techniques were used to sample painted surfaces, there was concern that some lead-
based paint in the home might go undetected. To minimize the probability of missing lead-based paint in a
home, technicians looked for and tested any painted surface in the dwelling unit deemed likely to contain
lead. If the first test failed to find lead, a second surface was tested. Due to practical constraints, the
technicians sometimes were limited to searching for lead-based paint in areas of the dwelling unit entered
during the course of the inspection and did not wander throughout the dwelling unit looking for lead-based
paint. A similar set of one or two purposive samples was taken for the exterior wall.
3-4
-------
Common Areas. Common areas were inspected in multifamily residences. If an interior common
hall existed, two dust samples were taken: one sample was taken from just outside the dwelling unit and
one just inside the main entrance to the building. If the building had common rooms such as a mail room,
laundry room, community room, etc., one was randomly selected and was inspected according to the
protocol used for the wet and dry rooms. Finally, playgrounds were inspected. Each piece of playground
equipment was quantified, described, and tested via XRF. Three soil samples were taken from the
playgrounds.
3 J Environmental Sampling Protocols
Paint Sampling. The lead-based paint testing was accomplished by MAP/XRF devices that
estimated the lead content of the paint.15 The MAP/XRF was equipped with a full-intensity 40-millicurie
Co57 source. The protocol called for the technician to place the MAP/XRFs scanner against the sampled
component, hold it stationary and flat against the surface throughout a 60-second reading, then record the
reading from the MAP/XRFs console.
Selection of MAP/XRF Equipment There were two primary rationales for choosing the
MAP/XRF to test for lead in paint The spectrum analyzer MAP/XRFs were used in preference over
direct-read XRFs because the National Institute of Standards and Technology (MIST) had determined them
to be more accurate and precise than the direct-reading XRFs used in earlier surveys.16'17 The MAP/XRF
was used in preference to taking paint scraping samples because the survey was conducted in occupied
dwellings where it was not feasible to take scrapings for laboratory analysis.
The spectrum analyzer MAP/XRF devices still had limitations. In particular, the MAP/XRF
generally produced readings mat were systematically different from the amount of lead in paint being
tested. The direction and magnitude of the differences were related to the substrate material and the lead
loading in the paint Furthermore, the precision of the readings depended on the substrate. On metallic
substrates, the MAP/XRFs readings were systematically higher man the true lead loadings, for all
observed levels of lead. On concrete, brick, and other related substrates, the MAP/XRF had great
difficulty detecting low to moderate levels of lead in paint, and there was much variation in readings.
Dust Sampling. The dust collection protocol employed a portable vacuum pump fitted with a length
of tubing, a fitter cassette, and a collection nozzle inserted over the filter cassette. The protocol specified
laying down a template that outlined an area of one square foot. The technician vacuumed the area in
overlapping passes, first left to right over the entire area and then top and bottom. Care was taken to hold
the nozzle level to the surface and to move the nozzle at a steady rate. Next, the template was moved over
one foot and the process repeated until four square feet were vacuumed. This protocol required about four
minutes. For areas which could not accommodate the template, such as window wells, the entire area was
vacuumed and dimensions of the area were recorded.
Constdentiaii was given to ic»ipiiig ample* of paint for laboratory analysis. Laboratory analysts ii more precise and accurate than In situ
MAP/XRF. However, it require damaging painted furftcei in people'i homo. It was felt that the gain measurement precision and accuracy would
be more than offid by eflfocd of a very large refusal rate.
16McKnignt, Mary E.; Byrd, W. Eric; and Roberts, Wfflard E. (May \99Q).MtasurtngLtadConcmtrattoiu tn Paint Uting a Portable Sptctnur
AnafyttrX-RayFtuorescmcfDcvIc* (NISTIR W90-650), US. Department of Commerce, National Institute of Standards and Technology.
17McKmght, Mary E.; Byrd, W. Eric; and Roberts, Willard E. and Lagergren, Eric S. (December 1989). Mtthods for Measuring Ltod
Concentration in Point Films (NISTIR 89-4209). US. TV|»"«'""« of Commerce, National Institute of Standards and Technology.
3-5
-------
Soil Sampling. The soil sampling protocol called far the field technicians to use a soil corer with
plunger to collect soil. The sample plug was expelled from ihe corer into a plastic bag. For each sampled
location, the technician drew three soil plugs from the ground, the first as close as possible to the targeted
sample she, and the other two twenty indies to either side of the first The protocol defined the soil sample
for each location as the blended composite of the three soil plugs.
3.4 Quality Assurance
Quality assurance was an integral component of the study design and execution. This section
summarizes the quality assurance procedures employed in the survey.
3.4.1 Measures to Enhance Response Rates
In the national survey, as any survey, maximizing response rates is critical to minimi ring bias and
maintaining the representativeness of the data. Maintaining participation at the point of the home
inspections was particularly important because of the investment of effort that was required to reach that
point For private housing, efforts to ensure participation began with in-person visits to homes for
screening purposes. In addition to the personal contact initiated by the interviewer, a letter from HUD
explaining the survey and verifying its legitimacy was left with the dwelling unit respondent
After dwelling units were selected from among the eligible dwelling units, telephone calls were
placed to the residents. The interviewers were trained to explain the survey and answer any questions or
concerns the resident might have. At times convenient to the respondents, the telephone interviewer
scheduled appointments for inspections to occur five to ten days after the phone call. Minimizing the delay
between the call and the inspection reduced the opportunity for the respondents to reconsider their
agreement to participate. The residents were informed at the time of this call that they would receive a $50
incentive for allowing the home inspection.
3.4.2 Quality Assurance for MAP/XRF Data
Extensive quality assurance (QA) procedures were developed for use of the MAP/XRF in the
national survey.
Eight "standards'1 were developed to conduct quality assurance readings. A "standard" consisted of
a small, 3" x 4" plastic sheet (i.e., shim) painted with paint containing a known level of lead and placed
tightly on a piece of background material (called a substrate). NIST manufactured the shims used hi the
survey. The substrates of interest were four common dwelling unit building materials: wood, drywall,
steel, and cement. All substrates used in the construction of standards had a smooth surface that allowed
the shim to be tightly and evenly affixed. Two different shims were used. One shim contained 0.6 rag/cm2
of lead, the other 2.99 mg/cm2 The four substrates and two shims were layered in combinations to make
eight test standards.
Baseline Readings. Upon receipt from the manufacturer, a quality assurance technician performed
eight 60-second readings on each of the eight standards. This produced a baseline of 64 readings of
different lead loadings on different substrate materials for each MAP/XRF used in the study. The testing
helped establish that each MAP was ready to be used in the field.
3-6
-------
Close-out Readings. The above procedures were repeated in full when a MAP/XRF was retired
from the field, and before returning it to the manufacturer.
In-Field Readings. In a scaled-down procedure, one reading was taken on each standard for a total
of eight readings every day that the MAP/XRF was in the field and in use. This practice had the additional
benefit of helping detect machine drift and malfunction. Technicians were instructed to call their
supervisor if readings varied from original baseline readings by over 30 percent. The procedure also helped
detect if batteries had become low and needed replacement. The in-field QA readings were not used by the
technicians to adjust the MAP/XRF readings in the field. Rather, the raw MAP/XRF readings were
recorded. During the data analysis, the QA readings were used to develop adjusted MAP/XRF readings.
3.4.3 Field Quality Assurance Measures
Prevention of Contamination of Dwelling Units and Samples. There was concern that the
inspection team might bring lead into the home. To help minimize this problem, each team member put
paper slippers over his or her shoes before entering the unit Team members wore rubber gloves during the
inspection. The technician discarded the gloves after he took each dust and soil sample, replacing them
with new ones. He replaced the dust sampling vacuum nozzle with a clean one after each sample. The
technician also cleaned the corer after each soil sample.
Quality Assurance for Field Equipment and Supplies. Efforts were taken to ensure that the
equipment or supplies themselves did not introduce lead contamination. Wet wipes used to clean soil
sampling equipment were tested for lead. Shavings from the soil sampling equipment, including the painted
handle, were tested. Particular caution was observed in testing supplies that technicians were expected to
purchase while in the field. Several brands of plastic bags and wet wipes were tested to ensure that
commonly available brands would not introduce any lead contamination.
3.5 Limitations of the Study Design
This research design reflects an explicit boundary of the survey. All audiences of the national
survey results should be aware it was never the survey's objective for the testing of sampled paint surfaces
within homes to be sufficient to prove or disprove if any specific home had a lead hazard, or the potential
for such. The data collection was designed to provide the supporting basis for national estimates, not to
determine definitively whether a particular house had lead-based paint present.
The statistical sample was specifically designed to permit the development of valid national
prevalence estimates, even with limited data on any particular dwelling. Given the ratio of sampled surfaces
to the total number of painted surfaces in a home, it is readily conceivable that lead-based paint could be
present in a given home but not detected as part of a statistically based sampling procedure. Conversely,
the testing protocol could have produced a very high lead reading where mere was, in fact, no current
hazard (e.g., lead-based paint was present, but was totally enclosed or covered with epoxy or acrylic).
Extensive lead testing designed to say with certainty whether a lead hazard existed in each unit was
beyond the scope of the survey and would have substantially increased costs and effort levels. The
increased time spent in each unit may have reduced the overall response rate, with a resulting decrease in
the quality of the overall data, rather man the increased quality mat might be expected from such increased
effort.
3-7
-------
4. SOURCES OF ERROR IN THE NATIONAL SURVEY DATA
In order to assess the utility of the national survey data for lead research and policy development, an
evaluation of the quality of the data is needed. Data quality is determined by the processes and procedures
that generated the dak including sample design and selection, response rates, field and laboratory
measurements.
One primary objective of the national survey was to estimate the nationwide percentage of dwelling
units with lead-based paint A dwelling unit is classified as having lead-based paint if any surface within
the dwelling unit has an average paint lead concentration greater than 1.0 mg/sq cm. The determination of
whether the average paint lead concentration across a surface was greater man 1.0 mg/sq cm was subject
to two sources of error: systematic differences between the MAP/XRF readings and the actual paint lead
concentrations; and misclassifications of surfaces due to random variability in the MAP/XRF readings.
The classification of a dwelling unit as having lead-based paint was subject to several additional error
sources, including: sampling error due to the random sampling of housing units; systematic differences
between the study respondents and non-respondents; sampling error due to the random selection of rooms
and painted components within rooms for the MAP/XRF measurements; and misclassifications of dwelling
units due to the incomplete sampling of rooms. Failure to take these sources of error into account in the
data analysis can produce biased prevalence estimates and overestimates of the precision in the estimates.
The difference between an MAP/XRF measurement and the actual lead concentration on the surface
being measured is called measurement error. The presence of measurement error does not mean that the
measurement was incorrectly made, but only that there was a non-zero difference (usually assumed to be
random) between the MAP/XRF measurement and the actual concentration. If the average of the
measurement errors is zero then the measurements are said to be unbiased and individual measurements
were sometimes higher and sometimes lower than the actual lead concentrations. If the average of the
measurement errors is other man zero, the measurements are said to be biased.
This chapter presents a brief discussion of these issues. Both sampling issues and MAP/XRF
measurement errors are included. A detailed discussion is presented in Appendix n, Chapter 3.
4.1 Response Rates and Potential for Non-Response Bias
An analysis of response rates was conducted to estimate the potential impact of national survey
nonresponse on the estimated prevalence of lead-based paint in housing. The analysis looked at the
relationship between response rates and factors such as geographic location, ethnicity, and economic
characteristics (e.g., rent, home value, and income) that could potentially influence response rates. In
addition, the analysis studied the association between inspected housing units in the same or nearby census
blocks with respect to the presence or absence of lead in paint, dust, and soil.
4.1.1 Private Housing
Table 4-1 displays contact, response, and eligibility rates at the national level. The eligibility rate
applies only to the screener stage, when in-person interviews were conducted to screen homes for eligibility
in the survey. Contact rates were calculated for bom the screener and telephone interview stages (when
respondents were asked to permit an inspection of then- home), whereas response rates were calculated for
all stages. The definition of a completed cosem Table 4-1 varies with the stage of data collection. A case
was considered completed at the screener stage if an interview was completed, whether or not the
respondent was eligible. The screener response rate thus measures the rate at which a screening interview
reaches its logical conclusion (eligible or ineligible), and did not terminate due to other causes (refusals,
4-1
-------
etc.). At the telephone interview stage, only eligible respondents who completed interviews and scheduled
appointments were considered completes. Finally, a case was considered complete at the inspection stage if
an inspection was completed.
Table 4-1
National Response, Contact, And Eligibility Rates At Each
Data Collection Stage
Data Collection Stage
S creener
Telephone Interview
Inspection
Contact Rate
77%
83%
Response Rate
63%
55%
90%
Eligibility Rate
89%
Notes:
The Contact Rate is the number nf mfitprtfi) hPUSCholds divided by the m-SCOpe mgnbCT attempted
Hie screencr Eligibility Rate is the number of eligible cases divided by the number of completed screening
interviews.
The Response Rate is the number of completed cases divided by the in-scope number completed.
The ethnic and economic distributions of housing units in the national survey were compared to the
corresponding distributions in the 1987 American Housing Survey performed by the Census Bureau and
HUD. No significant differences were formed with respect to ethnicity, building age or family income.
There were a few instances of significant differences: The national survey had fewer very low rent (< $200
per month) homes than expected, fewer high market value homes (> $150,000) than expected and more
homes in the South than expected. There is a strong positive association between inspected housing units in
the same census block with respect to the presence or absence of lead in paint, dust, and soil. That is, if
one house in a census block has lead-based paint, other houses in the same census block are also likely to
have it These findings suggest that the potential bias from nonresponding housing units is likely to be
small. Therefore, the estimates of lead-based paint prevalence were not modified to adjust for non-
response.
4.1.2
Public Housing
For the public housing component of the national survey, a sample of 110 projects from a national
frame of public housing projects was drawn according to the design described in Chapter 3. The survey
was completed in 97 of the sampled projects, indicating negligible potential for non-response bias in the
sample of public housing projects.
However, analysis of the sampling of housing units within the public housing projects showed that
there was an apparent over-representation of vacant housing units. The public housing survey design
called for randomly selecting one housing unit from each sampled project This design was difficult to
implement in the field because a number of the public housing project managers apparently felt it was more
convenient to permit the field team to sample vacant housing units. Consequently, there were 44 vacant
public hanging unite in the sample of 97 units. Although a comparison of the prevalence of lead in paint
between vacant and occupied housing units showed no significant differences, the impact on other target
parameters is unknown. For example, it would be difficult to estimate the percentage of public housing
units with lead-based paint and children under age seven because so many vacant apartments made up the
4-2
-------
sample. Furthermore, dust samples collected in vacant apartments probably do not represent samples
collected in occupied apartments. Because the public housing sample sizes were small and the occupied
housing unit sizes were even smaller, definitive conclusions resulting in national estimates of dust lead
levels are not possible.
4.2 MAP/XRF Measurement Bias
As described in Chapter 3, to assure a continual check of the accuracy and performance of the
MAP/XRF analyzers during the national survey, daily validation measurements were made. Statistical
analyses were performed on the validation data to estimate the precision and accuracy of the MAP/XRFs
readings, and to relate them to the substrate, concentration level, and age of the Co37 source. Equations
were developed to relate the lead concentrations on the substrates to the MAP/XRF readings. A different
equation was developed for each combination of machine and substrate. These equations were inverted to
generate equations to statistically calibrate the MAP/XRF readings.
This analysis of the validation data led to the conclusion mat the MAP/XRF devices used in the
National Survey generally produced readings that were systematically different from the actual amount of
lead in the paint. The direction and magnitude of the systematic differences were related to the substrate
material, the lead concentration in the paint, and, to a lesser extent, the age of the Co57 source. The exact
nature of the relationships varied from one individual MAP/XRF to another, sometimes significantly.
Furthermore, the precision of the readings depended on the substrate. Specifically, the MAP/XRF tended
to under-estimate low levels of lead in paint on wood, drywall, and plaster and over-estimate high levels of
lead. Also, the MAP/XRFs readings were systematically greater than the true lead loadings on steel and
other metallic substrates and it had great difficulty detecting low or moderate levels of lead in paint on
concrete, brick, and other related substrates.
Censoring of the MAP/XRF Readings
Although the cause of biased readings can come from several factors, an important factor in mis
study is censoring. The MAP/XRF instruments used in the national survey were designed to never report a
reading below zero even though readings below zero were otherwise possible. These MAP/XRF readings,
which would have been below zero but which were recorded as zero, are referred to as censored
MAP/XRF readings and contribute to bias. In the survey, censoring affected primarily the low readings on
concrete substrates. Since mis group of readings comprised only a small proportion of all MAP/XRF
readings, the effect of censoring on the national estimates is not significant
4.3 Classification Error Due to MAP/XRF Measurement Variation
Random variation in the MAP/XRF measurements can lead to incorrect classification of a surface
as having or not having lead-based paint The MAP/XRF measurement on a surface with paint lead
concentration less than 1.0 mg/sq cm may be greater than 1.0, or vice versa, resulting in a
misclassification. If the number of surfaces which are misclassified as having lead based paint is different
than the number of surfaces misclassified as not having lead based paint, the estimated percentage of
surfaces in a dwelling unit with lead-based paint will be biased. This is called classification bias. Note
that classification biases can exist even when the MAP/XRF measurements are unbiased. Estimates of the
percentage of surfaces within a dwelling unit with lead-based paint can be either low or high by as much as
eight percent Estimates of the proportion of surfaces nationally which have lead-based paint will have
less man one percent bias.
4-3
-------
4.4 Classification Error Due to Dwelling Unit Sampling Procedures
Within each dwelling unit, MAP/XRF measurements were made on randomly selected surfaces
within one randomly selected wet room and one randomly selected dry room. However, the classification of
a dwelling unit as having lead-based paint depends on the lead concentration on all surfaces within a
dwelling unit Using only the MAP/XRF measurements on the measured surfaces, a dwelling unit might be
misclassificd as not having lead-based paint when it actually did, because the surfaces with lead-based
paint were not among those selected for measurement
To help accurately predict the prevalence of housing units with lead-based paint and thus, estimate
the misclassification rate, it was necessary to develop a mathematical simulation model based on data from
the survey and other sources. The model was used to statistically remove the classification biases induced
by incomplete sampling of rooms and surfaces within a dwelling unit and by measurement variation.
Extending the results from the measured surfaces to all surfaces in the unit was based on: data on the
number of rooms in the unit; data on the number of surfaces per room; and assumptions about the
relationship of the lead concentrations on unmeasured surfaces to those on the sampled and measured
surfaces. The classification of dwelling units as having or not having lead-based paint can be biased
downward by 10 to 25 percentage points when only a few surfaces within the dwelling unit are measured.
4.5 Impact on the National Estimates of Prevalence
All statistical finding* on the lead hazard in homes reported elsewhere in this report incorporate the
results of the analyses of this chapter. That is, the raw MAP/XRF readings have been corrected for
measurement bias and censoring, and the misclassifications due to measurement variation and sampling
within dwelling units have been corrected. Chapter 2 presents the national estimate (83 percent) of
prevalence of lead-based paint in privately-owned housing. Without the corrections of mis chapter, the
estimate would have been 74 percent
4-4
-------
5. CONCLUSIONS
5.1 Major Conclusions
This section presents a brief discussion of the major conclusions to be derived from the experience
of the national survey.
5.1.1 Study Findings
Lead-based paint is widespread in housing. An estimated 64 million (± 7 million) homes, 83 percent
(± 9%) of the privately owned housing units built before 1980, have lead-based paint somewhere in the
building. Twelve million (± 1 million) of these homes are occupied by families with children under the age
of seven years old. An estimated 49 million privately owned homes have lead-based paint in their interiors.
There are no significant differences in the prevalence of lead-based paint by type of housing, market value
of the home, amount of rent payment, household income, or geographic region.
Thirteen million homes - 17 percent of the pre-1980 stock - have dust lead levels in excess of the
federal guidelines, regardless of whether or not they have lead-based paint However, excessive dust lead
levels are associated with the presence of damaged lead-based paint. Fourteen million homes, 19 percent of
the pre-1980 housing stock, have more than five square feet of damaged lead-based paint Nearly half of
them (47 percent) have excessive dust lead levels.
Although a large majority of pre-1980 homes have lead-based paint, most of them have relatively
small areas of surfaces covered with it. The average privately-owned housing unit with lead-based paint
has an estimated 601 square feet of it on interior surfaces and 869 square feet (per unit) on exterior
surfaces. Over half of the leaded paint is on walls, ceilings, and floors. The amounts of lead-based paint
per housing unit vary with the age of the dwelling unit. Pre-1940 units have, on average, about three times
as much lead-based paint as units built between 1960 and 1979.
Lead paint is even more widespread in public housing; 86 percent (± 8%) of all pre-1980 public
housing family units have lead-based paint somewhere in the building. While most public housing units
have some lead-based paint, most of them have small amounts of it. The average public housing unit with
lead-based paint has an estimated 367 square feet on interior surfaces and 133 square feet (per unit) on
exterior surfaces. Most of the interior lead-based paint is on walls, while very little of the exterior walls are
painted.
5.1.2 Impact of Measurement Error and Lead Loading Variation on the Data Analysis
The data analyses and reports of findings should incorporate instrument and laboratory
measurement error. The MAF/XRF instrument produced readings with measurement errors. They were
systematically different from the actual lead concentrations in the painted surfaces and had random
variation. Similarly, the laboratory protocols used to measure the lead in dust and soil had measurement
errors. These measurement errors can induce systematic errors in the estimated extent of the lead hazard
from paint, dust and soil, and can also result in underestimates of the uncertainty in the estimated hazard.
Therefore, the field procedures must provide for the collection of the QA data necessary to estimate the
measurement errors and their impact on the study findings. Further, the data analyses must explicitly
estimate and correct for the impact of the measurement errors.
The data analyses and reports of findings should take account of the inherent variation in the
painted surfaces. In the national survey, it was not possible to test for the lead content of every painted
surface in every sampled housing unit Consequently, surfaces were sampled. This sampling protocol was
5-1
-------
designed to control the project costs and respondent burden by controlling the amount of time required for
an inspection. However, it did not adequately provide for the estimation of variation in lead concentrations
within surfaces, between surfaces in the same room, and between rooms. These sources of variation need
to be addressed to produce accurate estimates of die uncertainty in the national estimates. It is therefore
recommended that multiple MAP/XRF readings be taken at randomly selected locations on a subset of the
selected surfaces; at two or more components of the same type in the same room; and in two wet rooms and
two dry rooms. Although it may not be feasible to do this in all homes, a subset should be selected for
these additional readings.
5.13 Use of the Spectrum Analyzer MAP/XRF
In contrast to the HUD Interim Guidelines, substrate correction was a necessary step in the
accurate determination of the presence and amount of lead-based paint on surfaces, when using the
MAP/XRF. As discussed in Chapter 4, in the national survey, the MAP/XRF generally produced readings
that were systematically different from the amount of lead in the paint being tested. Consequently,
substrate corrections are needed to obtain accurate measurements of lead loadings. There are two possible
ways to do this:
1. Take frequent validation readings and analytically correct the readings using methods as
described in Appendix n. Readings need to be done on three or more different shims,
however, not just two shims as in the national survey. With only two shims, only a linear
model can be used to correct the MAP/XRF readings. It is possible, however, that a non-
linear model better describes the relationship between the readings and the actual lead
concentrations in the painted surfaces.
2. Perform substrate corrections in the field. The HUD Interim Guidelines describe substrate
correction procedures appropriate for direct reading MAP/XRFs, not the spectrum analyzer
MAP/XRF.
5.2 Additional Conclusions and Recommendations
The main purpose of this section is to identify selected lessons learned during the conduct of the
national survey and to develop recommendations for future field operations. The objective of the following
recommendations is to improve the representativeness and statistical validity of the data, e.g., develop
methods to enhance respondent participation. Additional recommendations are presented in Appendix D,
Chapter 4.
Applicability of Recommendations
The national survey operated under constraints that are not typical of other lead-based paint studies.
Therefore, some of the following recommendations would not be appropriate under other research designs.
For example, most lead studies are conducted in a single locality, alleviating the problems associated with
moving teams and equipment around the nation under a tight schedule. The "nits in the national survey
were randomly selected from the national population of dwelling units; most other studies limit themselves
to a smaller, targeted population of units which have some underlying rationale for being included, e.g.,
they were FHA-mortgaged properties or they were in a targeted area of an inner city. A rationale for each
recommendation is provided and the reader must use his or her judgment in determining if the
recommendation is appropriate for another design.
All listing and screening should be completed before beginning unit inspections. Completion of
listing and screening before beginning unit inspections means that the full sample frame can be developed
5-2
-------
before sampling is begun. Because of the tight field schedule of the national survey, it was necessary to
begin inspections before listing and screening were completed in some counties. Although it was possible
to take samples from completed counties based on estimates of the final national distribution of eligible
housing, it was necessary after screening was totally completed to adjust the final sample to the actual
distribution. There were two ways to accomplish this: retroactively adjust the sample in the counties which
had been previously sampled; or sample the remaining counties to balance the sample to the national
distribution. Although the former approach yields a better sample statistically, it would have created costs
that were deemed to outweigh the gain in statistical power. The survey adopted the second course,
adjusting the sample in the counties that had not yet been sampled.
When dealing with PHAs, interviewers should closely coordinate with the PHA representative.
Coordination with the PHA management staff is imperative to effective sampling and inspection of public
housing. To ensure that all responsible and affected parties are involved in the process, it is advisable to
establish a liaison with both a PHA headquarters representative and the manager of the specific housing
project. Often teams met the PHA representative at the PHA office and then went to the site. The PHA
staff "smoothed" the way many times. In neighborhoods where the inspection teams fete unsafe, PHA staff
provided a vital service as escorts. Because the PHA staff are busy, every effort must be made to
accommodate their involvement For example, PHA offices often are not located close to the dwelling unit
Inspection teams should schedule their time to allow mem to meet the PHA staff at the PHA office and go
from there to the dwelling unit The PHA representatives typically tried to inspect as many units on each
trip as possible. This consideration, combined with frequent unexpected delays, suggests mat the
inspection team should not schedule other inspections too soon after tine PHA work, to avoid schedule
conflicts. Upon meeting a PHA representative in the field, the inspection team should not assume mat he or
she has been fully briefed on the study or the team's mission. The interviewer needs to go over the
objectives and procedures of the survey and the importance of the PHA dwelling unit inspections.
For public housing dwelling unit sampling, several additional weeks should be allowed for
dwelling unit sample selection hi large PHAs. The larger the PHA, the longer time it took to select the
dwelling units. The reasons varied, e.g., the contact was on vacation, there was difficulty finding the
contact, the contact needed authorization, etc. Without exception, selection of PHA dwelling units in the
five largest PHAs in the sample took several weeks longer than it did for the 25 other PHAs.
Field staff should be trained to respond to anyone's questions concerning what they are doing
and what the study is about Do not attempt to use local police or similar agencies, or the regional
HUD office, to substantiate or verify presence or activities, hi the national survey each field person
carried an ID badge, copies of the license that allowed him to handle the MAP/XRF and its radioactive
source, endorsement/introductory letters from survey and agency principals, the 800 number for the survey
coordinator, and a number for the Washington HUD representative. Field staff were questioned on several
occasions by police officers, building managers, or nearby residents. The documentation materials listed
above explained and substantiated the activities of the inspection team and proved invaluable. Although it
seems like a good idea, contacting local police in advance and then referring questioners to the police can
lead to a bigger problem; i.e., the police contacts inadvertently denying knowledge of the survey. If the
person who gets the call at the police station has not been apprised of the study, he or she will deny
knowing anything about it, even though the police department has been informed of the activities. There is
also a problem with overlapping jurisdictions. The study might notify the county sheriff, although the
inquiry comes into the state police.
The analytic model concerning dust analysis and pathways needs to be fully specified before an
adequate dust sampling protocol can be designed. Dust sampling and test result analysis are the subject
of continuing investigations. Insufficient dust was a recurring problem in the field. Outside of insufficient
dust, there were few field problems encountered with dust sampling. A traveling team operates under two
major constraints it must carry its own equipment and it must limit its time in the dwelling unit to a
5-3
-------
reasonable length. Getting more dust by using a bigger, more powerful vacuum or vacuuming a larger area
were limited options. The vacuum carried by the survey team weighed ten pounds, plus tubing, nozzles,
and extension cords. The survey found that it took four minutes to completely collect a dust sample
(vacuuming and re-vacuuming a 4-square-foot area as specified in the dust sampling protocol). A bigger
vacuum or more time spent vacuuming were not reasonable in the context of the national survey. At one
point in the procedure design stage it was suggested that in-field technicians evaluate the quantity of dust
collected in a cassette. They would collect more dust in the cassette if they determined there was
insufficient dust This suggestion proved to be impractical. First, inspectors could not reliably determine
by visual inspection if enough dust had been collected. Second, movement of the cassette while attached to
the vacuum could cause dust to fell out of the cassette. Third, allowing the technician individual discretion
to collect samples would lead to inconsistencies in the procedure and findings Last, the amount of time
spent vacuuming had to be limited to keep the visit to a reasonable length. Therefore, this suggestion was
not implemented in the survey. A final suggested approach to increasing dust sample yield was "wet wipe"
testing, used in conjunction with or following vacuuming, to pick up leftover dust. This technique would
not have improved effective yield, because wet wipe test results could not be compared or added in any
meaningful way to vacuumed dust sampling results. Static electricity posed a problem to the inspection
team. Dust would cling to the vacuum nozzle, the edges of the template, etc. Efforts were made to build
the template out of a material that did not accumulate static electricity but the phenomenon still occurred.
The MAP/XKF reading can be taken at any convenient place on a sampled architectural
component; valid sampling does not require readings of a randomly selected spot on the component
This «»liniinat«»g readings being taken in hard-to-reach locations, such as corners. The original design
called for sampling a random location on the surface of a component for testing. Though mis is subject to
further verification, it appears that components with common paint history produced similar MAP/XRF
readings. Another stage of sampling would significantly burden the inspection process. Correct
assessment of "common paint history" is more pivotal here.
All MAP/XRF readings should be recorded on paper by the inspector. The MAP/XRF used in
the national survey was programmed to store the spectrum results in memory. Serious problems were
encountered, though, when the memory was in use. The memory would fill up rapidly and the MAP/XRF
would stop operating. The equipment had to be turned of£ and men back on, in order to restore operation.
All memory was lost in the process. The need to link readings to a location necessitated recording of a
certain amount of data on paper in any event, so that recording MAP/XRF readings on paper involved little
extra effort. The equipment problem aside, using the MAP/XRFs on-board electronic storage necessitates
some procedures for downloading memory contents during the field period. In practice it did not prove
reasonable or practical to return the MAP/XRF to the survey operations office periodically for
downloading. Teaching field staff how to download the memory and transmit it to the field operations
office would have meant providing them with a properly-configured PC and modem in the field and
providing PC training. Adding responsibilities and equipment to the technicians' load was deemed
inadvisable.
5-4
------- |