Supplement

2013 American Housing Survey Sample Design and Weighting
SAMPLE SIZE
The 2013 American Housing Survey (AHS) data are from a sample of housing units interviewed
between May and September 2013. The same basic sample of national housing units is
interviewed every two years until a new sample is selected. The sample of supplemental
metropolitan units follows a longer interviewing cycle. The U.S. Census Bureau updates the
sample by adding newly constructed housing units and units discovered through coverage
improvement efforts.
For the 2013 AHS, approximately 167,500 sample housing units were selected for interview.
This sample includes 98,800 supplemental metropolitan sample units from 25 metropolitan
statistical areas (MSAs). Some units that were reduced from the AHS sample in 2007 have been
reinstated, and additional expansions to the sample have been made such as the inclusion of
approximately 5,100 subsidized housing units and the addition of eight MSAs that have never
been interviewed for the American Housing Survey – Metropolitan Sample (AHS-MS).
SAMPLE SELECTION
The Census Bureau has interviewed the current sample of housing units since 1985. First, the
United States was divided into areas made up of counties or groups of counties and independent
cities known as primary sampling units (PSUs). A sample of these PSUs was selected. Then, a
sample of housing units was selected within these PSUs.
Selection of sample areas. The sample for AHS is spread over 394 PSUs. These PSUs cover
878 counties and independent cities with coverage in all 50 states and the District of Columbia.
If there were over 100,000 housing units in a PSU at the time of selection, the PSU is known as a
self-representing PSU because it was removed from the probability sampling operation. It was
in sample with certainty. The sample from the PSU represents only that PSU. There are 170
self-representing PSUs.
The Census Bureau grouped the remaining PSUs and selected one PSU per group, proportional
to the number of housing units in the PSU, to represent all PSUs in the group. These selected
PSUs are referred to as nonself-representing PSUs. The sample nonself-representing PSUs for
AHS are a subsample of the Current Population Survey's (CPS) sample areas based on the 1980
census.
Selection of sample housing units. The AHS sample consists of the following types of units in
the sampled PSUs:



Housing units selected from the 1980 census
New construction in areas requiring building permits
1





Housing units missed in the 1980 census
Other housing units added since the 1980 census
Housing units selected from the 2000 census

Housing units selected from the 1980 census. The U.S. Census Bureau picked a systematic
sample so every unit had a 1 in 2,148 chance of being included in the AHS.
In areas where addresses are complete (at least 96 percent of units having a house number and
street name) and permits are required for new construction, housing units receiving 1980 census
long-form questionnaires were sorted by the following items:







PSU
Central city, urbanized area, urban outside urbanized area, rural
Owner, renter, vacant for rent, vacant for sale, other types of vacants
Number of rooms
Value of housing unit or gross rent
Manufactured/mobile home or not a manufactured/mobile home

In areas where addresses are not complete or permits are not required for new construction, land
areas were sorted using a formula incorporating the following items:













PSU
Central city, urbanized area, urban outside urbanized area, rural
Median value of housing unit
Number of children under 6 years old
Number of elderly people
Number of owner-occupied housing units
Number of manufactured/mobile homes
Number of housing units lacking some plumbing
Number of owner-occupied housing units whose value is below $45,000
Number of renter-occupied housing units with rent below $200
Number of Black and Hispanic people
Number of 1-room housing units

New construction in areas requiring building permits. In areas that require building permits for
new construction, the Census Bureau selected a sample of permits. These permits do not cover
manufactured/mobile homes or conversion of older buildings (i.e., warehouses) to residential
use.
Housing units missed in the 1980 census. The Census Bureau conducted a special study which
identified units at addresses missed or inadequately defined in the 1980 census. A sample of
these identified units was selected.

2

Other housing units added since the 1980 census. If extra units are added in buildings or
manufactured/mobile home parks where AHS already has sample units, a sample of these extra
units was selected. To find when whole buildings are built (in addition to building permits,
mentioned above) or are converted from nonresidential to residential use, the Census Bureau
listed all residential buildings in a sample of areas around the country, found any additional
buildings, and selected a sample of their units.
Housing units selected from the 2000 census. The following adjustments were made to the AHS
sample in 2005 by adding certain types of units selected from the 2000 census:


A new sample of manufactured/mobile homes was selected from the 2000 census
in an attempt to improve coverage of manufactured/mobile homes built between
1980 and 2000. These units were selected at the same rate as the 1980-based
units. One-half of this sample was included in the 2013 interviewing and, as a
result, one-half of the 1980-based sample was not included.



In an attempt to improve coverage of the elderly, a sample of special living units
was selected from the 2000 census. The Census Bureau purchased address lists of
special living facilities from two vendors and supplemented these lists with
facility addresses from various special living facility websites. These addresses
were matched to the 2000 census to identify 2000 census housing units in special
living facilities from which the resulting sample of these special living units was
drawn. Because the universe was so small, a double sample of units was selected.
Before interviewing, the resulting sample was screened to confirm eligibility.

Subsidized Housing Units. A sample of subsidized housing units was selected from address lists
provided by HUD in an attempt to improve coverage of housing units receiving rent subsidies.
The Census Bureau received address lists from HUD in 2010. These lists include the Public and
Indian Housing Information Center (PIC), Tenant Rental Assistance Certification System
(TRACS), and Home Investment Partnership Program (HOME) address lists. The lists were
matched to the sample PSUs to enable the Census Bureau to select housing units receiving HUD
subsidies.
Supplemental Metropolitan Sample
The Census Bureau initially grouped the housing units enumerated in the 1990 (2000) Census of
Population and Housing into census blocks and assigned these blocks to either the unit/group
quarters frame or the area frame, as follows:
1.

Blocks located in an area that issued permits for new construction were assigned to the
unit/group quarters frame.

2.

All other blocks were assigned to the area frame.

3

The unit/group quarters frame was then split into the unit frame and the group quarters frame by
removing all group quarters and placing them in a separate frame.
All housing units that were built after the 1990 (2000) census in areas where construction of new
housing units was monitored by building permits were placed into a separate frame, called the
permit frame.
Sampling operations for all frames were performed separately within a designated group of
counties in each state. Prior to the supplemental metropolitan sample selection, records selected
by other Census Bureau surveys were removed from each of the frames to avoid having the same
housing unit in sample for more than one survey. The Census Bureau selected the supplemental
metropolitan sample from the remaining records.
The following adjustments were made to the supplemental metropolitan sample in 2005 by
adding certain types of units selected from the 2000 census. These units will remain in the 2013
sample:
*

*

A new sample of manufactured/mobile homes was selected from the 2000 census
in an attempt to improve coverage of manufactured/mobile homes built between
1990 and 2000. One-half of this sample was included in the 2005 interviewing
and, as a result, one-half of the 1990-design sample was not included.
In an attempt to improve coverage of the elderly, a sample of special living units
was selected from the 2000 census.

NATIONAL ESTIMATION
Each housing unit in the AHS sample represents itself and over 2,000 other units. The exact
number it represents is its "weight." The weight was calculated in six steps. The purpose of
these steps is to minimize both sampling errors and errors from incomplete data. The result of
these steps is also to force consistency with some major categories of data in other Census
Bureau surveys, so figures on these categories do not actually depend on the AHS sample, but on
the other surveys.
1.

Basic weight. The Census Bureau assigned each unit a weight to reflect its probability of
selection. Generally, this weight is 2,148. Some exceptions are the special living sample
units, and the basic weights in the 30 metropolitan areas.

2.

Noninterview adjustment. An adjustment was made for refusals and occupied units
where no one was home. The calculations for this adjustment do not include units the
Census Bureau could not locate. The earlier weight was multiplied by the following
factor:
Interviewed units + Units not interviewed
Interviewed units

4

It is assumed the units missed are similar in some ways to the units interviewed for AHS.
This adjustment is done separately for groups defined by cross-classifying the following
data items if prior year data for the indicated items is available:










Nine census divisions
1990 central city, suburb, or nonmetropolitan
1990 urban or rural
Manufactured/mobile home or not a manufactured/mobile home
In a special living facility or not in a special living facility (only if the housing
unit is not a manufactured/mobile home)
Owner/for sale or renter/for rent
Number of units in structure*
Number of rooms*
Occupied, vacant year round, or seasonal/migratory vacant*

(* If known from a previous survey; otherwise, the Census Bureau substituted whether or
not units were drawn from building permits for these items.)
For seasonal/migratory vacants and year round vacants other than those for rent or for
sale, units were cross-classified only by census region and central
city/suburb/nonmetropolitan.
3.

PSU adjustment. The Census Bureau adjusted for differences that existed in 1990
between the number of 1990 census housing units estimated from the AHS sample of
nonself-representing (NSR) PSUs and the 1990 census counts outside the
self-representing PSUs. The earlier weight was multiplied by the following factor:

1990 census housing units in all areas that
could have been chosen as nonself-representing PSUs
1990 census housing units estimated from the
AHS sample of nonself-representing PSUs

This adjustment is done separately for groups defined by cross-classifying:

Nine census divisions

Owner, renter, or vacant

1990 central city, suburb, or nonmetropolitan

1990 urban or rural

Hispanic or non-Hispanic householder (only in South and West regions)

Black or non-Black householder (only in South region)

5

4.

New construction adjustment. The Census Bureau adjusted for known deficiencies in
sampling new construction by multiplying the earlier weight by the following factor:
Independent estimate
AHS sample estimate
This adjustment is done separately for groups defined by cross-classifying:





Nine census divisions
Manufactured/mobile home or not a manufactured/mobile home
Number of units in structure
Year built (pre-1980 and five-year categories after 1980 as shown in the
publication)

Independent estimates are based on the Census Bureau's Survey of Construction and
Survey of Manufactured Home Placements. Note that final AHS figures for the
categories above are not really based on the AHS sample findings, but on the independent
sources.
5.

Demographic adjustment. Comparability among the surveys was ensured by multiplying
the earlier weight by the following factor:
Independent estimate
AHS sample estimate
This adjustment is done in two steps for occupied units. First, the factors were computed
and applied for the Hispanic or non-Hispanic groups defined by cross-classifying:






Nine census divisions
Owner or renter
Hispanic or non-Hispanic householder
Husband-wife, other male householder, or other female householder
Age of householder

Next, the demographic adjustment is repeated with the same cells, except classified by
the Black or non-Black groups, rather than the Hispanic or non-Hispanic groups.
Vacant for sale, vacant for rent, other year round vacant, and seasonal/migratory vacant
units were cross-classified only by the four census regions and 2000 central city, suburb,
or nonmetropolitan area.
The percentage of occupied and vacant units was based on the AHS itself. The
distribution within occupied and vacant units are from the Census Bureau's Current
Population Survey for occupied units and from the Housing Vacancy Survey for vacant
units. The grand total number of all housing units in the United States is based on the
6

2000 census adjusted to account for new and lost units. Note that final AHS figures for
the categories above are not really based on the AHS sample findings, but on the
independent sources.
Repetitions. The new construction and demographic adjustments were repeated to help match
both sets of independent estimates simultaneously. These adjustments were repeated until every
cell's factor is between 0.98 and 1.02 or the change in each factor from one repetition to the next
is less than 0.015.
Small cells. In each step of weighting, many items were cross-classified; so some cells may have
few cases. When a cell is too small (less than 30 cases for the noninterview adjustment or less
than 50 cases for the demographic adjustment) or the adjustment factor is too extreme (greater
than 1.5 for the noninterview adjustment or outside a range of 0.5 to 2.0 for the demographic
adjustment) the Census Bureau combined the cell with one or more other cells that are similar in
most respects. Cells for the PSU adjustment or the new construction adjustment were not
combined.

SUPPLEMENTAL METROPOLITAN SAMPLE
SAMPLE AREA
The 2013 American Housing Survey’s supplemental metropolitan sample provides information
on 25 metropolitan areas interviewed as part of the AHS. The data from these 25 areas will be
used in creating the national estimates as well as metro level estimates.
In all of the historic AHS-MS metropolitan areas, the supplemental sample units include units
selected from the 1980, 1990, and 2000 censuses and any new construction since the censuses.
These metropolitan areas are consistent with the June 2003 Office of Management and Budget
(OMB) definitions of the metropolitan statistical area (MSA) as a result of the following sample
adjustments:
1.

2.

Counties/Minor Civil Divisions (MCDs) were added or dropped so that the definition of
each metropolitan area in sample was consistent with the final 2003 OMB definition of
the metropolitan area and sample was selected in these added areas.
The sample in the counties/MCDs in the previous definitions that were also in these new
definitions (i.e., continuing counties/MCDs) were adjusted to maintain an overall sample
size of approximately 4,500 and in some cases it was replaced by new sample for
confidentiality reasons.

In order to improve coverage for the eight new MSAs, the Census Bureau sampled units from the
July 2012 Master Address File for all counties in these MSAs, as given in the June 2003 OMB
definitions. Prior to sample selection, units already included in the AHS-N were removed from

7

the frame. Once the frame was created, units were grouped using the following items for each
MSA:






County
New construction from the 2010 census / 2010 census units
Tenure (Owner, Renter, Vacant)
Mobile homes
Single / Multi-unit dwellings

SAMPLE SIZE
The supplemental sample of 25 metropolitan areas includes about 98,800 units. When
aggregated with the 17,200 AHS-N cases in the AHS-MS areas, this yields approximately 4,500
from each of most MSAs.
Sample Housing Units
1.

2.

The sample from continuing counties consists of the following:
*

Interviews in the previous survey

*

Type A noninterviews (that is, units eligible to be interviewed) or Type B
noninterviews (that is, units not eligible for interview at the time of the survey but
which could become eligible in the future) in the previous survey.

*

New construction housing units selected from a listing of new residential
construction building permits issued since the previous survey. This sample
represented the housing units built in permit-issuing areas since the previous
survey

*

Housing units added since the previous survey in sample blocks from the
nonpermit universe. This sample represented the housing built in nonpermitissuing areas since the previous survey

*

2000 census manufactured/mobile homes

*

2000 census special living units

*

2000 census housing units replacing current sample housing units for
confidentiality reasons

The sample from new counties consists of the following:
*

2000 census housing units
8

*

New construction housing units selected from a listing of new residential
construction building permits issued since the 2000 census. This sample
represented the housing units built in permit-issuing areas since the 2000 census.

*

Housing units added since the 2000 census in sample blocks from the nonpermit
universe. This sample represented the housing built in nonpermit-issuing areas
since the 2000 census.

ESTIMATION FOR SUPPLEMENTAL METROPOLITIAN SAMPLE
Each housing unit in the AHS sample represents itself and other units. The exact number it
represents is its "weight." The weight was calculated in several steps.
1.
2.
3.

Basic weight. The Census Bureau assigned each unit a weight to reflect its probability of
selection.
Sample adjustment. An adjustment is made to the units remaining after the reduction.
Type A noninterview adjustment. Type A noninterviews are occupied sample units for
which occupants:
*
*
*

were not home
refused to be interviewed
were unavailable for some other reason

The Census Bureau performed the Type A noninterview adjustment by cross-classifying
occupied housing units into various categories or cells on the basis of the following data
items:
*
*
*
*
*
*
*

Central city/balance
Frame
Tenure (i.e., owner or renter)
Type of unit (i.e., mobile home, special living, non-mobile home or special living)
Rent
Value
Number of rooms

Within a given cell, the Type A noninterview adjustment factor was equal to the
following ratio:
Weighted Count of interviewed housing units +
Weighted Count of Type A noninterviewed housing units
Weighted count of interviewed housing units
9

4.

Mobile home ratio estimation. To adjust for undercoverage of manufactured/mobile
homes, the Census Bureau applied the following ratio estimation procedure in all areas:
Independent estimate of manufactured/mobile homes for the corresponding
geographic subdivision of metropolitan area
Sample estimate of manufactured/mobile homes for the corresponding
geographic subdivision of metropolitan area

The numerator of this ratio was determined using data from the 1980 census, the 1990
census, and the 2000 census. The denominator was obtained using the existing weight of
supplemental metropolitan sample mobile home units (i.e., the product of the basic
weight and the Type A noninterview adjustment factor).
5.

Independent total housing unit ratio estimation. For the ratio estimation procedure
described below, each metropolitan area was subdivided into geographic areas consisting
of individual counties or a combination of counties.
To lower the undercoverage of non-mobile housing units, the Census Bureau applied the
following ratio estimation procedure in all areas:
Independent estimate of total housing inventory (excluding mobile homes)
for the corresponding geographic subdivision of metropolitan area
Sample estimate of total housing inventory (excluding mobile homes)
for the corresponding geographic subdivision of metropolitan area

The numerator of this ratio was determined by a model consisting of the following
components:

a. Census 2000 Housing Units. The 2000 census counts of housing units are updated
each year through the Geographic Update System to Support Intercensal Estimates to
reflect boundary updates from the Boundary and Annexation Survey, Count Question
Resolution actions, and administrative revisions.
b. Estimated Residential Construction since April 1, 2000. This component is
calculated through a formula involving counts of new residential construction in nonpermit issuing areas since April 1, 2000 plus counts of residential building permits
that resulted in the construction of new units times a factor of 0.98 (since two percent
of all building permits never result in the actual construction of a housing unit ).
c. Estimated New Mobile Home Placements. The Census Bureau derives estimates for
manufactured/mobile homes by allocating state manufactured/mobile home shipment
data to subcounty areas based on the subcounty area’s share of state
manufactured/mobile homes in the 2000 census.
d. Estimated Housing Loss. The yearly estimates of housing unit loss are based on data
derived from the 1997-2003 American Housing Survey national sample (AHS-N).
10

The following three types of AHS noninterviews were considered to represent
permanent loss of a housing unit:
*
*
*

Type B-16 – Interior exposed to the elements
Type C-30 – Demolished or disaster loss
Type C-31 – House or Manufactured/Mobile Home moved

Housing unit loss rates based on these non-interview types were then developed for
housing units based on structure type and age of structure.
e. Final State and County Housing Unit Estimates. The housing unit estimates at the
subcounty level are summed to obtain county level housing unit estimates, which are
then summed to produce state level housing unit estimates.
For a more detailed description of the determination of these numbers, see
http://www.census.gov/popest/topics/methodology/2005_hu_meth.pdf. The denominator
was obtained using the product of the basic weight and the weighting factors of
supplemental metropolitan sample units, excluding manufactured/mobile homes.
The computed ratio estimation factors were then applied to all appropriate housing units in the
corresponding geographic area of each metropolitan area, and the resulting product was used as
the final weight for tabulation purposes.
The ratio estimation procedure reduced the sampling error for most statistics below what would
have been obtained by simply weighting the results of the sample by the inverse of the
probability of selection. Since the housing population of the sample differed somewhat by
chance from the metropolitan area as a whole, one can expect that the sample housing
population, or different portions of it, is brought into agreement with known good estimates of
the metropolitan area housing population.

11