2021 NYCHVS - Sample Design and Error Estimation

2017 New York City
Housing and Vacancy Survey

Sample Design, Weighting, and Error
Estimation

Updated: August 1, 2018

U.S. Census Bureau, Department of Commerce
New York City Department of Housing Preservation & Development

This page is intentionally blank.

Table of Contents
1.

Overview ................................................................................................................................. 1

2.

Sample Design ......................................................................................................................... 1

3.

4.

5.

2.1.

Eligible Universe………………………………………………………………………...1

2.2

Sampling Frames .............................................................................................................. 2

2.3.

Sample Selection by Frame .............................................................................................. 4

2.4.

Sample Size ...................................................................................................................... 6

Weighting ................................................................................................................................ 8
3.1.

Base Weight ..................................................................................................................... 8

3.2.

Nonresponse Adjustment ................................................................................................. 9

3.3.

Ratio Adjustment Factors for Housing Unit Weights .................................................... 10

3.4.

Ratio Adjustment Factors For Person Weights .............................................................. 12

Nonsampling Errors ............................................................................................................... 12
4.1.

Coverage Error ............................................................................................................... 13

4.2.

Nonresponse Error.......................................................................................................... 14

4.3.

Measurement Error from Missing Responses to Questions ........................................... 14

4.4.

Quality Validity Error .................................................................................................... 15

4.5.

Processing Error ............................................................................................................. 15

4.6.

Additional Considerations .............................................................................................. 16

Sampling Errors ..................................................................................................................... 16
5.1.

Sampling Error for Counts ............................................................................................. 16

5.2.

Sampling Error for Percentages ..................................................................................... 21

5.3.

Sampling Error for Differences ...................................................................................... 22

5.4.

Sampling Error for Medians ........................................................................................... 24

5.5.

Sampling Error for Means .............................................................................................. 26

6.

Replicate Weights .................................................................................................................. 28

7.

References ............................................................................................................................. 30

Appendixes
Appendix A: Example of Ratio Adjustments
Appendix B: List of Variables Being Imputed for 2017 NYCHVS
Appendix C: Housing Unit Characteristics Associated With GVF Parameters

This page is intentionally blank.

1. Overview
The purpose of this document is to describe the sample design, weighting, and error estimation
for the 2017 New York City Housing and Vacancy Survey (NYCHVS). The NYCHVS is
sponsored by the New York City Department of Housing Preservation and Development (HPD)
and conducted by the US Census Bureau.
The City of New York is required by law to conduct a survey periodically to determine if rent
regulations should be continued. A primary tool in this decision is the "vacant available for
rent" rate, which is defined as the ratio of the vacant available for rent units to the total number
of renter occupied and vacant available for rent units for the entire city. The NYCHVS measures
rental and homeowner vacancy rates, as well as various household and person characteristics.
The design requires the standard error of the estimate of the vacant available for rent rate for the
entire city be no more than one-fourth of 1 percent, if the actual rate was 3 percent.

2. Sample Design
The NYCHVS is a longitudinal survey that is conducted about every three years. The main
sample of the survey is selected every decade, and additional new units are selected in each
subsequent NYCHVS cycle. For the decade 2010-2020, the NYCHVS was conducted in 2011,
2014, and 2017. The main sample was selected after the 2010 Decennial Census, and then
additional sample units were selected in 2011, 2014, and 2017.
2.1. Eligible Universe
The universe of interest for the NYCHVS consists of the residential housing units (HUs)
located within the five boroughs of New York City (Bronx, Brooklyn, Manhattan,
Queens, and Staten Island). The principal exclusions are living quarters at locations that
are classified as group quarters. These include:










Transient hotels and motels (those that are less than 75% residential),
Correctional facilities,
Mental health institutions,
Hospitals,
Military installations,
Convents, monasteries, and rectories,
Shelters, group homes, communes, and halfway houses,
Home for the aged, disabled, homeless, or needy, and
Dormitories for students or workers.

Residential hotels and motels (those that are at least 75% residential); however, are
included in the survey.

2.2. Sampling Frames
The 2017 NYCHVS sample consists of housing units selected from the following four
sampling frames:
1.

Housing Units Included in the 2010 Census

This frame was created from the decennial census from both the Decennial HundredPercent File (DHF) and the Census Unedited File (CUF). Both the DHF and CUF are
created from the decennial response file, which contains all responses to the 2010
Census. The CUF contains the unedited individual responses to the 2010 Census
questionnaire, while the DHF contains the edited responses to the 2010 Census
questionnaire. The reason we used these files instead of Master Address File (MAF) is
the DHF and CUF had the most current data at the time of our sample selection.
The sample selected from this frame is referred to as the initial and main sample of the
2017 NYCHVS. This sample was included in all three cycles of NYCHVS: 2011, 2014,
and 2017. The sample housing units for the 2017 NYCHVS were initially selected from
this frame in 2011.
HPD obtained, prepared and provided to the Census Bureau three address list files, which
were used as additional sample frames for the 2017 NYCHVS:
2.

Housing Units Newly Constructed Since the 2010 Census

This list was based on New Construction data maintained by the NYC Department of
City Planning that integrates regular updates from the NYC Dept. of Buildings (DoB).
The New Construction list included only unique addresses with Final Certificates of
Occupancy issued by the DoB for newly constructed residential units since the last HVS
cycle (from December 2013 through November 2016). Addresses were cleaned and
prepared by HPD using the City’s Geo-Support system to validate house number and
street name addresses, to provide a valid Building Identification Number (BIN), BBL,
and number of residential units, to eliminate duplicates and invalid or pseudo-BINs.
Addresses without valid BINs were excluded.
The resulting frame contained units in buildings constructed after the 2010 Census. This
included units constructed prior to the 2011 survey (eligible for the 2011 survey), units
constructed since the 2011 survey but prior to the 2014 survey (eligible for the 2014
survey), as well as units constructed since the 2014 survey (eligible for the 2017 survey).
3.

Housing Units from Alterations and Conversions

To create this list, HPD obtained from DoB the latest available Alterations File, which
was current as of November 30, 2016 inclusive, and extracted listings for addresses
where DoB recorded a signed-off Alteration Permit dated between December 1, 2013 and
2

November 30, 2016. Only addresses where the records indicated a net gain in number of
residential units were extracted, along with the number of newly created units. By
processing the addresses using the City Planning Geo-Support system, each address was
linked to its unique BIN. Only addresses with valid BINs were retained; addresses with
duplicate or pseudo-BINs were dropped. The calendar year of the final sign-off was
recorded.
This frame is similar to the newly constructed frame (frame #2 listed above) and contains
residential units newly created in existing buildings since the 2010 Census. It includes
housing units constructed since the 2010 Census in preexisting buildings altered to create
more units (alterations) or converted from nonresidential use (conversions). This
includes units constructed prior to the 2011 survey (eligible for 2011 survey), units
constructed since the 2011 survey but prior to the 2014 survey (eligible for 2014 survey),
as well as units constructed since the 2014 survey (eligible for 2017 survey).
4.

Housing Units in Structures Owned by New York City (in rem)

This frame consisted of units in structures owned by New York City as of November
2016. The City owned these units because the owner failed to pay the real estate tax
and/or other charges on the property. HPD is the City agency responsible for
administering the inventory of in rem buildings. There were historically two different
administrative groups of in rem buildings, those centrally managed by the Division of
Property Management (DPM) and those in programs where responsibility for maintaining
and/or upgrading the buildings was delegated to different community organizations or
groups. The stock of in rem buildings is a dwindling universe. Over the years buildings
that were in various earlier in rem programs were transferred into other HPD programs
for rehabilitation or management. In order to maintain comparability of the sample
frames through the subsequent HVS cycles, subsequent or re-named programs still
containing “legacy” in rem buildings were identified and the addresses were consolidated
into a “DPM” list and a “community” list. Only buildings with zero residential units or
demolished building addresses were removed. The remaining “legacy” in rem addresses
were processed through the City Planning Geo-Support system, so that each address
could be linked to a unique BIN, BBL, etc. However, addresses that could not be linked
to a unique BIN were not removed, because HPD confirmed all addresses on the list as a
City-owned residential building.
The domain for this frame changes in every sampling cycle, since some new units are
added, and some units get sold and are no longer in rem. In the 2017 NYCHVS, the units
selected from this frame supplemented the in rem sample from previous sample years that
are still in rem.
The frame size for 2017 was 212 buildings containing roughly 1,700 HUs. The city
requires a sample size of approximately 600-700 units in the sample. Thus, these types
of housing units were oversampled to ensure a large enough sample for analysis of this
sub-universe.
3

The HUs of the in rem frame may also be part of the 2010 Census frame. We accounted
for the overlapping frames by adjusting the probability of selection of units in both
frames, and thereby their base weights as suggested by Lohr (2007, 2010).
2.3. Sample Selection by Frame
Within each of the four NYCHVS sampling frames, we selected clusters (groups of
housing units) of generally four housing units, with the exception of some in rem
buildings where we selected clusters of five.
Frame 1: Housing Units Included in the 2010 Census
The sample for this frame came from two different Census 2010 files – the DHF and the
CUF. The sample from the DHF was selected first and the CUF was second. To ensure
no HU was selected in both files, HUs in the CUF were removed if they were already in
the DHF.
Within this frame, we sorted housing units by








Borough,
Sub-borough,
Percent renter occupied in the block,
Tract,
Block number,
Basic street address, and
Unit designation.

We selected a systematic random sample of housing units across all boroughs from the
ordered frame. This frame included in rem units.
Frame 2: Housing Units Newly Constructed Since the 2010 Census
We selected units in this frame from Certificates of Occupancy (C of Os) issued between
April 2010 and November 2016. The list of C of Os was provided by HPD for each
survey cycle.
Sample units were selected for the 2011 survey from Certificates issued between April
2010 and November 2010. Additional sample units were selected for the 2014 survey
from Certificates issued between December 2010 and November 2013. Additional
sample units were selected for the 2017 survey from Certificates issued between
December 2013 and November 2016. Units selected from the C of O frame in 2011 and
2014 remained in the sample for 2017.
For the 2017 NYCHVS, we dropped from this frame all housing units that were also on
the 2010 Census frame, or previous 2011, 2014 C of O lists. We sorted the housing units
4

by borough, street name, and street number, and then selected a sample of housing units
within each borough. We listed each structure that contained a sample housing unit and
then identified the designated sample unit in the order in which the unit appeared on these
listings.
Frame 3: Housing Units from Alterations and Conversions
Housing units added to existing residential buildings (alterations) and housing units in
buildings converted from nonresidential use (conversions) were sampled for the 2017
survey. The selection process was conducted for the 2011 and 2014 surveys, and was
conducted again, using updated lists from HPD, for the 2017 survey. Addresses were
identified by HPD where residential units were created through alterations or conversions
with signed off permits between December 2013 and November 2016.
The list of alteration and conversion addresses was matched to the 2017 C of O frame list
for newly constructed buildings and to the 2010 Census on basic address. For matching
addresses, the unit counts were compared between the city’s alteration and conversion list
and the new construction C of O or Census 2010 list. If the alteration and conversion
listing for the address contained more units than the new construction C of O or the
Census list, it was considered an alteration and eligible for the alteration sample. If the
alteration and conversion listing for the address contained the same or fewer units than
the new construction C of O or the Census list, it was dropped from the alteration and
conversion frames because those units should have been accounted for in the C of O or
the Census list first. If the address did not match, the building was considered a
conversion and included in the conversion frame.
Within each frame, a sample of buildings was selected. The sampled buildings went
through a listing process from which sample units were identified. For the alterations
sample, a determination was made about which units were not included in the Census or
the new construction C of O file. These units were then eligible for the alterations sample.
For the buildings identified as conversions, all units listed were eligible for the conversion
sample.
Frame 4: Housing Units in Structures Owned by New York City (in rem)
The in rem frame is a special domain identified by HPD. The oversample of in rem HUs
is selected in each cycle of the survey from a frame that is updated in each cycle of the
survey. The main requirement is that the in rem universe is oversampled with a sample
size of 600-700 units each sample cycle.
This frame is the most complicated, because the in rem universe changes each sample
cycle (2011, 2014, and 2017); some units remained in the frame, some new units came in,
and some units dropped out. In 2011, a HU that is in rem could be selected into the
sample from two different frames: the 2010 Census frame and the 2011 in rem frame. For
2017, the third sample cycle of the NYCHVS in the decade, a HU that is in rem could be
selected into the sample from four different frames: the 2010 Census frame, and each of
5

the three in rem frames (2011 in rem universe, 2014 in rem universe, and 2017 in rem
universe).
If the sampled buildings selected in previous surveys, 2011 or 2014, did not drop out of
the sample for 2017, the sample units selected within those buildings will continue to
remain in the in rem sample for 2017. We only selected additional in rem units to replace
the lost in rem sample units from 2014.
We selected a probability-proportional-to-size sample of in rem buildings first, then
selected sample units within buildings. In this procedure, each building is assigned a
probability of selection based on the expected number of housing units in the building.
This probability is in direct proportion to this expected number of units. Thus, a building
with eight units has twice the probability of selection as a building that has four units.
First, we sorted the buildings by:




Borough,
Street name,
House number

We next selected a systematic random sample of buildings from the ordered frame. Then
we listed the individual units in each building, and last we selected a sample of units
within each sample building.
2.4. Sample Size
The total number of sample housing units selected for the 2017 NYCHVS was 19,020.
Table 2.3 provides the total number of selected housing units by borough, as well as the
breakdown of completed interviews and non-interviews.

Borough

Bronx
Brooklyn
Manhattan
Queens
Staten Island
Total

Table 2.3. Interview Activity for the 2017 NYCHVS
Unweighted
Weighted
Selected Completed
Type A
response
response
Interviews
Nonrate
rate
interviews
77%
76%
2,863
2,168
661
83%
83%
5,494
4,459
914
83%
82%
5,165
4,229
870
78%
78%
4,529
3,489
975
83%
83%
969
790
162
81%
80%
19,020
15,135
3,582

Type C
Noninterviews
34
121
66
65
17
303

Of these 19,020 total sampled housing units, 15,135 interviews were completed. The
NYCHVS classifies two types of non-interviews: Type A and Type C.
6

In 2017, there were 3,582 Type-A non-interviews. These include occupied housing units
where the occupants:




Refused to be interviewed,
Were not at home after repeated visits, or
Were unavailable for some other reason.

Type A non-interviews also include vacant units. In these cases, an interview was not
obtained if no informed respondent could be found after repeated visits.
There were an additional 303 Type-C non-interviews, which were not interviewed
because they no longer existed or were uninhabitable.
This classification produced an overall unweighted response rate of 81 percent (19,0203,582-303)/ (19,020-303) = (15,135/18,717). The response rate is calculated as the total
number of interviews divided by the total eligible sample, which can written as:

𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑅𝑎𝑡𝑒 =

𝑇𝑜𝑡𝑎𝑙 𝑆𝑎𝑚𝑝𝑙𝑒 − 𝑇𝑦𝑝𝑒 𝐴 𝑛𝑜𝑛𝑖𝑛𝑡𝑒𝑟𝑣𝑖𝑒𝑤𝑠 − 𝑇𝑦𝑝𝑒 𝐶 𝑛𝑜𝑛𝑖𝑛𝑡𝑒𝑟𝑣𝑖𝑒𝑤𝑠
𝑇𝑜𝑡𝑎𝑙 𝑆𝑎𝑚𝑝𝑙𝑒 − 𝑇𝑦𝑝𝑒 𝐶 𝑛𝑜𝑛𝑖𝑛𝑡𝑒𝑟𝑣𝑖𝑒𝑤𝑠

Note the response rate using the base weight is 80 percent. For calculating response
rates, all of the following must be answered to be considered as a completed interview:










Occupancy/vacancy status,
Year moved,
Coop/condo status,
Tenure,
Units in structure,
Interviewer observations of building condition items,
Contract rent,
Type of vacant unit, and
Asking rent

AND two of the following five items answered from the household roster for each
person:






Sex,
Age,
Relationship to householder,
Hispanic origin, and
Race.
7

If these criteria were not met, the sampled unit was classified as either a Type A or Type
C non-interview, following the definitions above.
Interviews started in January of 2017 and continued through May 2017 and survey
operation was conducted out of the Census Bureau’s New York Regional Office. We
hired over 400 field representatives who were tested and trained for this survey. In
addition, we filled other positions including field supervisors, automation clerks,
administrative clerks, recruiters, and an overall program coordinator. Most of the
respondent interviews were personal visits, but sometimes respondents did not agree to a
personal interview and in these cases a telephone interview was conducted.
For evaluation of falsification of interviews, a second interview was conducted of all
vacant units and five percent of all occupied units. The questions asked during the
reinterview included information about the previous field representatives (FR) that
collected data, the time, date, and length of that interview, tenure, and vacancy status.
In 2017, we did not conduct proxy or last resort interviews as in past surveys. Last resort
interviews in past surveys were conducted for reluctant respondents in which we
designated certain questions as essential and accepted an abbreviated questionnaire as
complete. Essential items in the past included tenure, rent, vacancy status, year moved,
demographic information, among other items. In the past, proxy interviews were
conducted when we were not able to conduct an interview, after numerous attempts.
Proxy interviews were interviews with a real estate agent, building manager, or someone
who knew something about the housing unit.

3. Weighting
In order to estimate housing unit and person characteristics based on the data we collected for
the 2017 NYCHVS, we calculated sample weights for each sample housing unit, and each
sample person. The final weight for each housing unit is the product of the following
weights and adjustments:
3.1. Base Weight
We determined the base weight as the reciprocal of the probability of selecting the unit.
Because in rem sample units and some census sample units were eligible for selection
from both the 2010 Census and the in rem frames, we adjusted the base weights to reflect
the fact that housing units had multiple chances of selection given our overlapping
frames.

8

3.2. Nonresponse Adjustment
We adjusted the base weight of each interviewed housing unit to account for the 3,582
eligible units that did not respond (Type A non-interviews). We applied this nonresponse
adjustment using a non-interview adjustment factor (NIAF). This factor was applied to
all interviewed housing units to account for Type-A non-interviews. The factor was
calculated using the following ratio:

𝑁𝐼𝐴𝐹 =

𝐼𝑛𝑡𝑒𝑟𝑣𝑖𝑒𝑤𝑠 + 𝑇𝑦𝑝𝑒 𝐴 𝑛𝑜𝑛𝑖𝑛𝑡𝑒𝑟𝑣𝑖𝑒𝑤𝑠
𝐼𝑛𝑡𝑒𝑟𝑣𝑖𝑒𝑤𝑠

We computed the factor separately for old construction (those are sample HUs selected
from the 2010 Census) and new construction (those are HUs added to the sample after the
2010 Census) housing units as follows:
Old Construction
For sample housing units selected from the 2010 Census frame, we computed the NIAF
separately by borough using the characteristics listed in Table 3.2. We used 2017
NYCHVS response data when available to determine the tenure and characteristics cell of
a unit. If the 2017 NYCHVS data were not available, we used 2014 NYCHVS response
data. If 2014 NYCHVS response data were not available, we used 2011 NYCHVS
response data. If 2011 NYCHVS response data were not available, we used the 2010
Census data. If the 2010 Census data were also not available, we treated it as a vacant
housing unit, and assigned it to the “HU without tenure or vacancy status” (unknown)
cell of the vacant housing unit table.
The process of determining the tenure and characteristics cell of a unit was different in
the 2014 survey cycle or prior. Prior to 2017, we used 2011 NYCHVS response data to
determine the tenure status. If 2011 NYCVHS response data were not available, we used
2010 Census data. If 2010 Census data were also not available, we used the current 2014
NYCHVS response data. Starting in 2017, we simplified this process by using the most
current values available.
Table 3.2 summarizes the variables used in combination to define cells of the NIAF
tables for old construction sample units.

9

Table 3.2 Variables Used to Define Nonresponse Adjustment Cells for Old Construction
NIAF
Variable
Values
Tables by
HU Type

RenterOccupied
Housing
Units

OwnerOccupied
Housing
Units

Vacant
Housing
Units

Borough
Monthly Rent

Number of
Rooms
Borough

Bronx
Brooklyn
< $100
$100-$199
$200-$299

Manhattan
Queens
$300-$399
$400-$499
$500-$599

Staten Island
$600-$699
$700-$999
$1,000+
Unknown

1-2, 3, 4-5, 6+, Unknown
Bronx
Manhattan
Brooklyn
Queens
< $25,000
$100,000-$149,999
$25,000-$49,999
$150,000-$199,999
$50,000-$74,999
$200,000-$249,999
$75,000-$99,999
$250,000-$299,999
1-5, 6-7, 8, 9+, Unknown

Staten Island

Bronx
Manhattan
Brooklyn
Queens
Vacancy Status Rented/sold/vacant
Seasonal/Occasional
for rent/vacant for
Migrant workers
sale
Other
Number of
1-2, 3, 4-5, 6+, Unknown
Rooms

Staten Island

Value of the
House

Number of
Rooms in the
Housing Units
Borough

$300,000-$399,999
$400,000-$499,999
$500,000+
Unknown

Unknown

New Construction, Alterations, and Conversions
For new construction units, alterations, and conversions, we computed the factor
separately using the year the segment was selected (2011, 2014, or 2017) and borough.
3.3. Ratio Adjustment Factors for Housing Unit Weights
We adjusted the housing unit weights using three ratio adjustments with the following
known totals:




The 2010 Census frame totals,
The in rem frame totals,
Housing unit totals produced by demographic analysis

For each ratio estimation procedure, we computed and applied factors separately by cells.
The factors were equal to the following ratio:
10

𝐾𝑛𝑜𝑤𝑛 𝑇𝑜𝑡𝑎𝑙𝑠
𝑁𝑌𝐶𝐻𝑉𝑆 𝑆𝑎𝑚𝑝𝑙𝑒 𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒
The denominators of the ratios are equal to the sum of the weights of housing units (or
persons) with all previous factors applied. Appendix A includes more information on the
ratio adjustment factors, and examples on how the process works.
The three ratio adjustments are described below:
1.

2010 Census Ratio Adjustment Factor (RAF)

This ratio adjustment accounted for differences between the 2010 Census counts and the
Census characteristics of the corresponding weighted sample counts. The factor reduces
the variability resulting from sampling the 2010 Census frame. We adjusted the weights
of all NYCHVS sample units selected from the 2010 Census frame, and computed the
factors separately by borough using the following 2010 Census characteristics. Table 3.3
summarizes the variables used in combination to define cells of the Census ratio
adjustment.
Table: 3.3 Variables Used to Define Census Ratio Adjustment Cells
Variable
Values

RAF Table by HU
Type

Renter-Occupied
Housing Units

Sub-borough

Number of Persons in
the Housing Units
Race and Ethnicity of
the Householder
Owner-Occupied
Housing Units

Vacant Housing
Units

Sub-borough

Number of Persons in
the Housing Units
Type of Vacancy

Borough

2.

Bronx: 1-10
Brooklyn: 1-18
Manhattan: 1-10
1, 2, 3-4, 5 or more

Queens: 1-14
Staten Island: 1-3

White (non-Hispanic)
African American (non-Hispanic)
Other (all remaining races)
Bronx: 1-10
Queens: 1-14
Brooklyn: 1-18
Staten Island: 1-3
Manhattan: 1-10
1, 2, 3-4, 5 or more
Vacant for rent
Vacant for sale
Rented/sold
Bronx
Brooklyn
Manhattan

Seasonal
Migrant
Other
Queens
Staten Island

In rem Ratio Adjustment

This procedure adjusts for known sampling variability in the in rem sample selection.
We adjusted the weights of all sample units selected from the in rem frame by borough
11

(five cells). We used the total number of units in each borough in the in rem frame as
control totals.
3.

2017 Housing Unit Ratio Adjustment

This procedure adjusted the 2017 NYCHVS sample estimate for sampling variability and
housing unit undercoverage (as described in Section 4.1.) by controlling the sample
estimate using independent estimates of 2017 total housing units. The independent
estimates were projected to 2017 based on 2010 Census housing unit totals. The
independent estimates were derived from the Census Bureau’s demographic population
estimates program and are used here to correct for the coverage error (for more
information, see Census, 2017a). We applied this ratio estimation procedure to all
interviewed housing units. We calculated the ratio adjustment factor for each of the
boroughs (five cells). The independent estimates were counts of the total number of
housing units in each of the boroughs at the time of the 2017 survey.

3.4.

Ratio Adjustment Factors For Person Weights

When calculating person weights, the final housing unit weight was used as the base
weight for each person, then we added a ratio adjustment to account for sampling
variability and known coverage deficiencies (as described in Section 4.1.) for persons
within interviewed households. We computed this factor within each borough by age,
race, Hispanic Origin, and sex (200 cells).


The numerator of the ratio equaled the independent estimate of 2017 total
persons for the cell minus the NYCHVS sample estimate of reference persons
and spouses or unmarried partners. The independent estimates were projected
based on 2010 Census person totals (Census, 2017a).



The denominator of the ratio equaled the NYCHVS sample estimate of
persons other than reference persons, spouses, or unmarried partners for the
cell. The person ratio estimate factor was applied only to the persons other
than reference persons, spouses, or unmarried partners.

The ratio estimation procedures, as well as the overall estimation procedure, reduced the
sampling error for most statistics in comparison to what would have been obtained by
simply weighting the sample by the base weight.

4. Nonsampling Errors
Since the statistics produced from this survey are estimates derived from a sample, they will
differ from the “true values” being estimated. There are two types of errors, which cause
12

estimates based on a sample survey to differ from the true value - sampling error and
nonsampling error.
If every housing unit in New York City were interviewed, the estimates of housing unit
characteristics could still differ from the true value (for example, the median contract rent).
In this instance, the difference is due solely to nonsampling errors. We attribute nonsampling
errors in sample surveys to many sources:








Deficiencies in the sampling frame (i.e., not all housing units are covered),
Inability to pick up all persons within sample households,
Inability to obtain information about all cases in the sample,
Definitional difficulties,
Differences in the interpretation of questions,
Inability or unwillingness to provide correct information on the part of the
respondents, and
Mistakes in recording, coding or keying the data obtained.

There are also other errors of collection, response, processing, coverage, and estimation for
missing data.
4.1. Coverage Error
Coverage errors arise from the failure to give some units in the target population any
chance of selection into the sample (undercoverage), or giving units more than one
chance of selection (overcoverage). To calculate the coverage, we used the sample base
weight, or the weight prior to any sample adjustments. The sample adjustments described
in Section 3, help to mitigate the undercoverage identified in this section.
The coverage rate is the ratio of the survey population or housing unit estimate of a group
or an area and the independent estimate (or the known totals). The undercoverage rate is
calculated as:
𝐾𝑛𝑜𝑤𝑛 𝑇𝑜𝑡𝑎𝑙𝑠
𝑈𝑛𝑑𝑒𝑟𝑐𝑜𝑣𝑒𝑟𝑎𝑔𝑒 = (
− 1) ∗ 100
𝑁𝑌𝐶𝐻𝑉𝑆 𝑆𝑎𝑚𝑝𝑙𝑒 𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒
Based on the Census population estimate for NYC in March 2017, in the 2017 NYCHVS,
we missed less than one percent of the housing units in the five boroughs. Overall, we
missed about twelve percent of the people in sample households. (See Table 4.1a)
Table 4.1a Overall Undercoverage for HUs and Persons
Estimated from
Known Total
Undercoverage
NYCHVS
Housing Units
3,469,240
3,489,271
0.58%
Persons
7,519,528
8,418,512
11.96%

13

The within-household undercoverage varied by age, race, sex and borough. Table 4.1b
gives the undercoverage of the various race-sex groups for the city as a whole:
Table 4.1b Undercoverage by Race/Ethnicity-Sex Group
Race/Ethnicity-Sex Group
Undercoverage
White & Other Females
2%
White & Other Males
3%
African American Females
18%
African American Males
22%
Asian Females
6%
Asian Males
4%
Hispanic Females
18%
Hispanic Males
29%
We adjusted for this undercoverage through the housing unit and person ratio adjustment
factors. These factors adjust the sample weights to population totals provided by the US
Census Bureau, the resulting final weight accounts for the undercoverage identified in
Table 4.1. For more information on the sample adjustment process see sections 3.3 and
3.4. NYCHVS data users do not have to take any additional steps to account for
coverage error.
4.2. Nonresponse Error
Some respondents refuse the interview or cannot be located. We mitigated the error due
to nonresponse by applying the non-interview adjustment factors into the weighting
process, as discussed in Section 3.2. NYCHVS data users do not have to take any
additional steps to account for nonresponse error.
4.3. Measurement Error from Missing Responses to Questions
Some respondents participate in an interview but refuse to answer questions or do not
know a particular answer. For many housing, demographic, and economic questions the
Census Bureau imputes missing responses. When imputing, we try to find households or
persons with similar characteristics to fill in missing data. For each imputation, records
are divided into 'donors' and 'recipients.'
For the demographic items we first try to impute based on other household
information. Every household must have some demographic information or it would be
made a Type C. It is rare that a household is missing all demographic information for
one item.
For imputing the housing items, we look for units with similar characteristics. For
example, when imputing contract rent, we find a unit with a similar year moved, year
14

built range, units in structure, and input control status (stabilized, public housing,
unregulated, etc.) and use that unit’s contract rent to impute the recipient's contract
rent. If no such unit can be found, we impute contract rent based on the median value for
units with the same input control status in the particular borough. In 2017, contract rent
was imputed in only four percent of the renter occupied units.
For economic items, we try to achieve the best possible match between donors and
recipients through a statistical match with key items. The items used for matching donors
and recipients are sex, race, ethnicity, age, relationship, education, worked last week,
temporarily absent, looking for work, year last worked, kind of business, type of
business, industry, occupation, weeks worked, hours worked, and rent/value. We use all
of these criteria to get the best statistical match possible.
All donors and recipients have the same borough, tenure, and either receive public
assistance or do not. Appendix B provides the list of variables being imputed.
Variables that can be used to determine imputation rates are in the public use files (PUF)
and are defined on the record layout. These variables are shown, beginning on page 23
for occupied units, page 33 for persons, and page 43 for vacant units. For example, using
these variables from the PUF, users can see we imputed electricity for 5.6 percent of
occupied units, we imputed age for 2.9 percent of all persons, and we imputed stories for
5.4 percent of vacant units.
The Census Bureau does not know how close the imputed values are to the actual values.
For other items, “not reported” is used as an answer category. NYCHVS data users do
not have to take any additional steps to account for nonresponse error.
4.4. Quality Validity Error
In order to design a survey question that accurately measures the constructs of interest,
the Census Bureau carefully tests each new survey question to ensure it is measuring the
construct of interest. While we have an English and Spanish questionnaire, sometimes
the respondent does not speak either one of these languages. In these cases, the interview
must be rescheduled so that a FR that speaks the same language as the respondent can
administer the interview. Although some respondents might misinterpret questions, the
Census Bureau does not have any additional information to estimate validity error rates.
NYCHVS data users do not have to take any additional steps to account for validity error.
4.5. Processing Error
The 2017 NYCHVS was administered using a paper instrument. This requires more
processing than most other Census surveys, which are completed using a computerized
survey instrument. For each interview, the survey data are keyed and verified by our

15

National Processing Center, and then transmitted to our programming area while the data
are reviewed and edited.
After the data are collected, errors that can be introduced include data capture errors, data
coding and classification errors, and data editing and imputation errors. The Census
Bureau carefully tests all aspects of the data capture, coding, classification, editing, and
imputation procedures. Although mistakes are possible, the Census Bureau believes they
are minimal. If a processing error is discovered, the Census Bureau will let NYCHVS
data users know and, in some cases, will publish revised estimates. NYCHVS data users
do not have to take any additional steps to account for processing error.
4.6. Additional Considerations
The NYCHVS is a longitudinal survey conducted every three years. Many NYCHVS
users compare current year NYCHVS with prior year estimates. Users should be aware
that HPD and the Census Bureau often make small changes to the text of various
questions between surveys. NYCHVS data users comparing estimates with prior year
surveys should consult the ‘Overview’ document on the NYCHVS website (See Census
2014 and Census 2017b).

5. Sampling Errors
Sampling error is a measure of how estimates from a sample vary from the actual value. By
the term "actual value" we mean the value we would have gotten had all housing units been
interviewed, under the same conditions, rather than only a sample.
Users of NYCHVS PUF can use replicate weights to estimate errors for any estimate. For
further information, see Section 6.
The Generalized Variance Functions (GVFs) are a convenient tool for quick and easy
estimation of sampling errors. The text below describes how to calculate sampling errors for
counts, percentages, differences, medians, and means using GVFs.
5.1. Sampling Error for Counts
Most published estimates from the NYCHVS reflect weighted counts of housing units.
The error from sampling for a weighted count is approximated using the following GVF
for estimating a 90-percent confidence interval:
1.645 ∗ √𝑎 ∗ 𝑋̂ + 𝑏 ∗ 𝑋̂ 2 ,
Where 𝑋̂ is the weighted sample estimate from the file, and a and b are GVF parameters
that vary depending on the characteristics being estimated.
16

Sets of GVF parameters have been computed for New York City as a whole, as well as
for each of the five boroughs. For 2017 NYCHVS, the GVF parameters are now
calculated using replicate weights, see Section 6 for more information on replicate
weights.
Table 5.1a contains GVF parameters for computing standard errors of housing unit
characteristics and Tables 5.1b and 5.1c contain the GVF parameters for computing
standard errors of person characteristics.
Housing Unit Characteristics
The parameters provided in Table 5.1a identify three sets of GVF parameters for housing
units of NYC and each of the five boroughs. Use Table 5.1a and Appendix Table C1 and
C2 to decipher which GVF parameters to use for a given housing unit characteristic.
Table C1 identifies the characteristics to be used with the second set of parameters.
Table C2 identifies the characteristics to be used with the third set of parameters. For a
given estimate, consider the geography first and then refer to tables C1 and C2. If the
characteristic can be matched to either table then use the parameters associated with that
table. The first column in Tables C1 and C2 lists the characteristics for which the tables
are to be applied. The second column lists the applicable subgroup (e.g. total occupied,
vacant for rent, etc.) If the estimate of interest matches to both the first and second
column of either table, use the corresponding GVF parameters. If the characteristic of
interest is not identified in either Table C1 or C2 then use the first set of GVF parameters.
Exhibit 5.1 illustrates how to correctly select the right set of GVF parameters for
calculating sampling errors.
For sub-borough estimates, the sub-borough is treated as a characteristic and can be
found on table C1 or C2 depending on the borough. Match the borough and the
characteristic to table C1 and C2 to determine which set of parameters to use.
Person Characteristics
The parameters provided in Table 5.1b and Table 5.1c identify seven sets of GVF
parameters for person estimates of NYC and each of the five boroughs. To help
determine which parameter set to use for a given person estimate, first consider the
geography then identify matching characteristics. If no characteristics can be matched to
the ones listed then use the parameters identified for “person characteristics not listed
above”. If multiple sets of characteristics can be matched then use the set of parameters
yielding the higher standard error.
For sub-borough estimates, find the parameters given for the borough and use the
parameter for the person characteristics “Borough and Sub-borough”.

17

Exhibit 5.1: Decision Tree on How to Determine Which Set of GVF Parameters to Use
Characteristic of
Interest

HU
Characteristic?

No

Use GVFs from Table
5.1b or 5.1c for
Person
Characteristics

Yes

Use 2nd set of GVF
parameters from
Table 5.1a

Yes

Use 3rd set of the
GVF parameters
from Table 5.1a

Yes

In Table C1?

No

In Table C2?

No

Use 1st set of GVF
parameters from
Table 5.1a

18

Table 5.1a The Value of a and b Parameters for Housing Unit Characteristics by Borough
Borough
HU Characteristics …
a
b
City Wide
… not listed in C1, C2
284.23
-0.000082
… listed in table C1
368.42
-0.000106
… listed in table C2
304.08
-0.000087
Bronx
… not listed in C1, C2
322.97
-0.000613
… listed in table C1
404.18
-0.000767
… listed in table C2
380.36
-0.000722
Brooklyn
… not listed in C1, C2
296.17
-0.000286
… listed in table C1
378.88
-0.000366
… listed in table C2
312.72
-0.000302
312.90
Manhattan
… not listed in C1, C2
-0.000354
396.23
… listed in table C1
-0.000449
385.20
… listed in table C2
-0.000437
297.82
Queens
… not listed in C1, C2
-0.000348
378.94
… listed in table C1
-0.000443
334.69
… listed in table C2
-0.000391
315.61
Staten Island
… not listed in C1, C2
-0.001746
373.75
… listed in table C1
-0.002071
469.29
… listed in table C2
-0.002604

City Wide

Borough

Table 5.1b The Value of a and b Parameters for Person Characteristics for City Wide
Person Characteristics
a
b
White and Other Race Ethnicity
544.38
-0.000034
Males
441.75
-0.000030
Female
459.51
-0.000050
Under 25 years old and other special characteristics1
396.74
0.000157
African Americans, American Indians or Native
607.77
0.000127
Alaskans
Borough and Sub-borough2
604.47
-0.000054
Person characteristics not listed above
518.85
-0.000021

1

Special characteristics include: retired, income less than $20,000, highest education level is H.S. diploma and not
enrolled in any other education, and self-employed.
2
Exclude total population in households. Use the set of GVF parameters for “characteristics of persons not listed
above” for these person characteristics.

19

Staten Island

Queens

Manhattan

Brooklyn

Bronx

Table 5.1c The Value of a and b Parameters for Person Characteristics by Borough
Borough
Person Characteristics
a
b
White and Other Race Ethnicity
576.83
-0.000168
Males
452.41
-0.000195
Female
478.43
-0.000139
3
Under 25 years old and other special characteristics
393.93
0.000705
African Americans, American Indians or Native Alaskans
564.87
0.000848
4
Borough and Sub-borough
643.66
-0.000138
Person characteristics not listed above
512.98
-0.000046
White and Other Race Ethnicity
511.72
0.00002
Males
432.17
0.000022
Female
426.81
-0.000109
3
Under 25 years old and other special characteristics
372.57
0.000429
African Americans, American Indians or Native Alaskans
583.45
0.000556
4
Borough and Sub-borough
596.27
-0.000048
Person characteristics not listed above
470.45
0.000000
White and Other Race Ethnicity
525.09
-0.000076
Males
427.18
-0.000225
Female
406.88
-0.000119
3
Under 25 years old and other special characteristics
342.23
0.001138
African Americans, American Indians or Native Alaskans
401.88
0.000832
4
Borough and Sub-borough
570.34
-0.000141
Person characteristics not listed above
445.07
-0.000062
White and Other Race Ethnicity
547.76
0.000001
Males
444.81
-0.000113
Female
479.12
-0.000194
Under 25 years old and other special characteristics3
440.31
0.000473
African Americans, American Indians or Native Alaskans
574.96
0.000290
Borough and Sub-borough4
640.30
-0.000117
Person characteristics not listed above
515.26
-0.000063
White and Other Race Ethnicity
492.36
-0.000106
Males
380.83
-0.000526
Female
436.90
-0.000571
Under 25 years old and other special characteristics3
356.57
0.001552
African Americans, American Indians or Native Alaskans
521.66
0.010359
4
Borough and Sub-borough
568.62
-0.000398
Person characteristics not listed above
455.15
-0.000157

3

Special characteristics include: retired, income less than $20,000, highest education level is H.S. diploma and not
enrolled in any other education, and self-employed.
4
Exclude total population in households. Use the set of GVF parameters for “characteristics of persons not listed
above” for these person characteristics.

20

The parameters in Table 5.1a, 5.1b, and 5.1c, citywide and for each borough, allow you
to compute a range of error such that there is a known probability of being correct if you
say the actual value is within the range. The error formulas are approximations to the
errors. They indicate the order of magnitude of the errors rather than the actual errors for
any specific characteristic. To construct the range, add and subtract the error computed
from the formulas to the estimate.
We will continue with an example using the equation we provided for estimating the
sample error for counts. In the 2017 NYCHVS, there are 22,537 vacant-for-rent units in
Brooklyn, that is A=22,537. To compute a 90-percent confidence interval, you would use
the first set of GVF parameters in Table 5.1a and you would compute the margin of error
as follows:
1.645 ∗ √(296.17 ∗ 22,537) + (−0.000286 ∗ 22,5372 ) = 4,203
Thus, there is a 90-percent chance you will be correct if you conclude the actual number
of vacant-for-rent units in Brooklyn is 22,537 plus or minus 4,203 or in the range 18,334
to 26,740.
If the estimate involves two characteristics from Tables 5.1a, 5.1b or 5.1c, use the set of
GVF parameters with the larger a parameter.
5.2. Sampling Error for Percentages
Any subgroup can be shown as a percentage of a larger group. The error from sampling
for a 90 percent confidence interval for this percentage is computed as:
𝑎 ∗ 𝑃 ∗ (100 − 𝑃)
1.645 ∗ √
𝐴
where:
a:
P:
A:

the parameter a from Table 5.1a, 5.1b or 5.1c,
is the percent you calculate, and
is the weighted denominator of the percent.

For example, there are 580,484 occupied home owner conventional housing units in New
York City and 130,487, or 22.5 percent, were built between 1947 and 1973. Use Table
5.1a for City Wide, together with Table C1 and C2 in Appendix C. Since the
characteristic (year building built) is listed in Table C2, the applicable subgroups for this
characteristic do not include occupied home owner conventional housing units, you must
use the first set of the parameters from Table 5.1a. To compute a 90-percent confidence
interval you would plug the following numbers into the above formula:
21

1.645 √

284.23 ∗ 22.5 ∗ 77.5
= 1.52
580,484

Thus, if you say that the actual percentage of occupied home owner conventional housing
units in New York City built between 1947 and 1973 is between 20.98 percent and 24.02
percent, there is a 90-percent chance you will be correct.
5.3. Sampling Error for Differences
People often ask whether two numbers are actually different. Two numbers from the
NYCHVS, for example, 21 and 34, or 34 percent and 55 percent, have a statistically
significant difference if their 90-percent confidence intervals do not overlap.
When 90-percent confidence intervals do overlap, numbers are still statistically different
if the result of subtracting one from the other is more than:
1.645 ∗ √𝜎12 + 𝜎22
Where:
𝜎1 :
𝜎2 :

the standard error for the first number
the standard error for the second number

This formula is quite accurate for (a) the difference between estimates of the same item in
two different areas or (b) the difference between separate and uncorrelated items in the
same area. If there is a high positive correlation between the two items, the formula will
overestimate the error. If there is a high negative correlation, the formula will
underestimate the error. The following illustration shows how to compute the error of a
difference.
There are 5,603 condominium housing units in Queens with 20 to 49 units in the building
and 7,605 condominium housing units in Queens with 50 to 99 units in the building.
Follow the steps in Table 5.3a to compute the 90-percent confidence interval for the
difference between those two numbers.

22

Table 5.3a Steps to Compute the 90% Confidence Interval for a Difference
Steps for Calculations
The Formula
Which set of GVF parameters should we use?
(since the characteristic of interest is units in Condo
building in Queens, and this matches to both
columns in Table C2 of Appendix C, use the third
set of the parameters for Queens from Table 5.1a)

An Example

a

334.69

b

-0.000391

How many total units in Queens with 20-49 units in
the building?

̂1
𝑋

5,603

What’s the estimated standard error of total units in
Queens with 20-49 units in the building?

𝜎1 = √𝑎 × 𝑋̂1 + 𝑏 × ̂𝑋12

√334.69 × 5,603 − 0.000391 × 5,6032
= 1,364

How many total units in Queens with 50-99 units in
the building?

̂2
𝑋

7,605

What’s the estimated standard error of total units in
Queens with 50-99 units in the building?

𝜎2 = √𝑎 × 𝑋̂2 + 𝑏 × ̂𝑋22

√334.69 × 7,605 − 0.000391 × 7,6052
= 1,588

̂2 − 𝑋
̂1
𝐷𝑖𝑓𝑓 = 𝑋

7,605-5,603 = 2,002

𝑀𝐸 = 1.645 ∗ √𝜎12 + 𝜎22

1.645 ∗ √1,3642 + 1,5882 = 2,612

𝐷𝑖𝑓𝑓 ± 𝑆𝐸

2,002 ± 2,612

What’s the difference of the two numbers you are
interested in?
What is the margin of error for a 90-percent
confidence interval for the difference?
The 90-percent confidence interval for the
difference is:

Thus, a 90-percent confidence interval of (-610, 4,614) includes zero. Therefore, the
difference between condominium housing units in Queens with 20 to 49 units and 50 to
99 units is not statistically significant.
Here, we demonstrate how to compare the same estimate of two boroughs. For example,
we want to know whether the estimated number of rent stabilized housing units in the
Bronx is different from the Manhattan estimate. Table 5.3b provides the steps that
compute the 90-percent confidence interval for the difference between those two
numbers.

23

Table 5.3b Steps to Compute the 90% Confidence Interval for a Difference
Steps for Calculations
The Formula
Which set of GVF parameters should we use?
(since the characteristic of interest is units with
stabilized rent in Bronx and Manhattan, and
this matches neither tables of Appendix C, use
the first set of the parameters for Bronx and
Manhattan from Table 5.1a)

Bronx:
a1
b1
Manhattan:
a2
b2

How many total units in Bronx are rent
stabilized?

An Example

Bronx:
322.97
-0.000613
Manhattan:
312.90
-0.000354

̂1
𝑋

233,502

What’s the estimated standard error of total
units in Bronx with stabilized rent?

𝜎1 = √𝑎1 × 𝑋̂1 + 𝑏1 × ̂𝑋12

√322.97 × 233,502 − 0.000613 × 233,5022
= 6,480

How many total units in Manhattan are rent
stabilized?

̂2
𝑋

249,000

What’s the estimated standard error of total
units in Manhattan with stabilized rent?

𝜎2 = √𝑎2 × 𝑋̂2 + 𝑏2 × ̂𝑋22

√312.9 × 249,000 − 0.000354 × 249,0002
= 7,481

̂2 − 𝑋
̂1
𝐷𝑖𝑓𝑓 = 𝑋

249,000 - 233,502 = 15,498

𝑀𝐸 = 1.645 ∗ √𝜎12 + 𝜎22

1.645 ∗ √6,4802 + 7,4812 = 16,281

𝐷𝑖𝑓𝑓 ± 𝑆𝐸

15,498 ± 16,281

What’s the difference of the two numbers you
are interested in?
What is the margin of error for a 90-percent
confidence interval for the difference?
The 90-percent confidence interval for the
difference is:

Thus, a 90-percent confidence interval of (-783, 31,779) includes zero, so we conclude
that the difference between rent stabilized housing units in the Bronx and Manhattan is
not statistically significant.

5.4.

Sampling Error for Medians

The median is the value 50-percent of the way through the distribution. Thus, 50-percent
of the total falls below and 50-percent falls above the median. Note that the median
presented in this example is the true median (i.e., computed by statistical package) not an
approximation. You can construct a confidence interval around the median by computing
the standard error on a 50-percent characteristic and then translating that into an interval
for the characteristic.
Steps to compute the sampling errors for medians:
1) First, get the estimated standard error of a 50-percent characteristic, using the
same formula for errors for percent (section 5.3), but substitute 50 for the P:
24

√

𝑎 ∗ 50 ∗ (100 − 50)
=σ
𝐴

2) Then, calculate the standard error from sampling for the median as:
(𝑈 − 𝐿) ∗

σ
= 𝑆𝐸𝑚𝑒𝑑𝑖𝑎𝑛
𝑝

where:
a:
A:
U-L:
σ:
p:

is parameter a from Table 5.1a, 5.1b or 5.1c.
is the total number of housing units from the distribution.
is the width of the interval that contains the median. U is the upper bound
of the interval, and L is the lower bound of the interval.
is the error for a 90-percent confidence interval for the 50-percent
characteristic.
is the percent of cases that fall in the interval containing the median.

3) Last, calculate a 90 percent confidence interval for the true median by adding and
subtracting to the median:
𝑀𝑒𝑑𝑖𝑎𝑛 ± 1.645 ∗ 𝑆𝐸𝑚𝑒𝑑𝑖𝑎𝑛
For example, the median household income for all occupied housing units in New York
City is $57,500. The number of occupied housing units in the distribution of household
income is presented in the Table 5.4a.
Table 5.4a: Distribution of Household Income
Household Income
Less Than $5,000/no income/loss
$5,000-$9,999
$10,000-$14,999
$15,000-$19,999
$20,000-$24,999
$25,000-$29,999
$30,000-$34,999
$35,000-$39,999
$40,000-$49,999
$50,000-$59,999
$60,000-$69,999
$70,000-$79,999
$80,000-$89,999
$90,000-$99,999
$100,000-$124,999
$125,000-$149,999
$150,000 or more
TOTAL

Number of HUs
166,827
139,239
161,977
155,689
156,732
131,104
128,896
122,756
216,410
201,798
185,237
148,787
138,940
120,038
262,477
164,428
508,620
3,109,955

25

Percent
5.36
4.48
5.21
5.01
5.04
4.22
4.14
3.95
6.96
6.49
5.96
4.78
4.47
3.86
8.44
5.29
16.35
100.00

Cumulative Percent
5.36
9.84
15.05
20.06
25.10
29.31
33.46
37.40
44.36
50.85
56.81
61.59
66.06
69.92
78.36
83.65
100.00

The error on a 50-percent characteristic based on 3,109,955 units is calculated as illustrated
in the Table 5.4b.
Table 5.4b. Steps to Compute the 90% Confidence Interval for a Median Household Income
Steps for Calculations
The Formula
An Example
How many total units is the median based on (in
thousands, exclude “not reported” and “don’t
know”)?
What’s the parameter a? (since household income
is not a characteristic listed on either Table C1
and C2 of the Appendix C, use the first set of
parameters for citywide from Table 5.1)
What is the estimated standard error of a 50percent characteristic with a base equaling the
total units?

σ=√

What are the end points of the category the
median is in?

A

3,109,955

a

284.23

𝑎(0.5)(1 − 0.5)
A

284.23(0.5)(1 − 0.5)
√
3,109,955
= 0.0048

U–L

$59,999.5 – $49,999.5

What is the width of this category (in dollars,
rooms, or whatever the item measures)?

W

$10,000

How many housing units are in this median
category?

B

201,798

What is the estimated proportion of the total units
falling in the category containing the sample
median?

𝑃=

Then the standard error from sampling for the
median is approximately:
The 90-percent confidence interval for the
median is:

𝐵
𝐴

𝑠𝑒𝑚𝑒𝑑𝑖𝑎𝑛 =

𝜎×𝑊
𝑃

𝑀𝑒𝑑𝑖𝑎𝑛 ± 1.645 × 𝑠𝑒𝑚𝑒𝑑𝑖𝑎𝑛

201,798
= 0.0649
3,109,955
0.0048 × $10,000
= $739.60
0.0649
$57,500 ± $1,217

Thus, there is a 90-percent chance that you will be correct if you conclude that the actual
median household income for all occupied housing units in New York City is between
$56,283 and $58,717.

5.5. Sampling Error for Means
The mean and the median usually differ. The mean is usually higher because it is
influenced more heavily than the median by very large values. Use the following
equation to calculate a 90-percent confidence interval of the mean:
(∑𝑛 𝑝𝑖 𝑥𝑖2 − (∑𝑛𝑖=1 𝑝𝑖 𝑥𝑖 )2 )
1.645 ∗ √ 𝑖=1
∗𝑎
𝑐
26

where:
pi: is the proportion of total households or persons from a distribution in the ith interval.
xi: is the midpoint of the ith interval (The midpoint of the open-ended interval is 2.5
times the lower limit).
c: is the total number of households or persons in the distribution (Subtract the number
of "not applicable" from the total to get c).
n: is the total number of intervals in the distribution.
a: is the parameter a from Table 5.1a, 5.1b or 5.1c.
For example, the mean (or average) household income of all occupied housing units in
New York City was $97,132 (compared to a median of $57,500). The distribution from
which the mean was computed is given in Table 5.5.
Table 5.5: Distribution of Household Income from the Mean
Household Income

Number of HUs

Less Than $5,000/no income/loss
$5,000-$9,999
$10,000-$14,999
$15,000-$19,999
$20,000-$24,999
$25,000-$29,999
$30,000-$34,999
$35,000-$39,999
$40,000-$49,999
$50,000-$59,999
$60,000-$69,999
$70,000-$79,999
$80,000-$89,999
$90,000-$99,999
$100,000-$124,999
$125,000-$149,999
$150,000 or more
Total

pi

xi

166,827
139,239
161,977
155,689
156,732
131,104
128,896
122,756
216,410
201,798
185,237
148,787
138,940
120,038
262,477
164,428
508,620

0.0536
0.0448
0.0521
0.0501
0.0504
0.0422
0.0414
0.0395
0.0696
0.0649
0.0596
0.0478
0.0447
0.0386
0.0844
0.0529
0.1635

3,109,955

1.000

$2,500
$7,500
$12,500
$17,500
$22,500
$27,500
$32,500
$37,500
$45,000
$55,000
$65,000
$75,000
$85,000
$95,000
$112,500
$137,500
$375,000

The error for a 90-percent confidence interval on the mean value is computed as follows:
26,772,828,750−(106,827.502 )

1.645 ∗ √

3,109,955

∗ 284.23 = $1,949

Thus, there is a 90-percent chance of being correct if you say the mean household income
of all occupied housing units in New York City is between $95,183 and $99,081.

27

6. Replicate Weights
New to the NYCHVS 2017 data files are replicate weights. These replicate weights provide the
data user an alternative method of producing variance estimates. Both the GVFs provided in this
document and replicate weights on the data file can be used to produce estimates of variance.
The GVFs provide a convenient way of producing a variance estimate while the use of replicate
weights would be a more specialized and technical way to estimate variance. Both are
acceptable approaches to variance estimation, and the method chosen would depend on the data
user’s familiarity with each method, access to statistical software, and data user preferences.
Variance estimation for surveys refers to the variation of an estimate due to selecting a sample
from the set of all possible samples for a given sample design. So to estimate the variance we
need multiple samples but we only observe one. Replication uses the single observed sample to
generate several replicate samples. These replicate samples can then be used to measure the
variation of the estimates. Replication allows small changes to a single probability sample to
create a set of replicate samples. This is done through subsets selected from the original sample
in a process that mimics the original sample design. Each sample replication (r) is then fully
weighted, using the same process as the original sample, to ensure each replicate sample, r,
represents the population of interest. This process forms the set of final replicate weights {𝑤𝑟 | r
= 1, … ,R}, this is similar to what is provided in the 2017 NYCHVS data file. Considering a
particular estimate of interest, each replicate weight, 𝑤𝑟 , can be used to create a replicate
estimate 𝜃̂𝑟 . The set of replicated estimates {𝜃̂𝑟 | r = 1, … , R} represents the variability, or
dispersion, of the estimate of interest under multiple samples of the population. Using the
replicated estimates together with the 2017 NYCHVS equation of replicate variance, data users
can calculate an estimated variance of an estimate of interest.
The 2017 NYCHVS uses a replicate variance estimator derived from a variance equation called
the successive differences estimator. This estimator was first introduced by Fay and Train
(1995) and then expanded for replication by Ash (2014). To calculate the variance of an
estimate, use the replication variance estimator:
80

4
2
𝑣̂(𝜃̂) =
∑(𝜃̂𝑟 − 𝜃̂0 )
80
𝑟=1

where ˆ is the weighted estimate of the statistic of interest; such as a total, median, mean,
proportion, regression coefficient, or log-odds ratio, using the weight for the full sample and ˆr is
the replicate estimate for replicate r of the same statistic using the replicate weights. 𝜃̂0 is the full
sample estimate. The value of 80 in 𝑣̂(𝜃̂) is the number of replicates used – NYCHVS uses 80
replicates.
There are two sets of replicate weights. One set of replicate weights is used for computing
standard errors of housing unit characteristics and the second set is used for computing standard
errors of person characteristics. This is similar to our GVF parameters, we have two different
28

sets of GVF parameters one set for housing characteristics and another set for person
characteristics.
To calculate a standard error, the measure of dispersion when parameter estimates are calculated
through repeated sampling from the population, obtain the square root of the variance estimate.
The following example illustrates how a statistic would be estimated, replicated, and combined
to form a variance estimate. We are going to estimate the variance using the 80 replicate weights
provided for the NYCHVS. Please note that for 2017 NYCHVS, the weights in Replicate 1
equals full sample weights, or the weight used to derive sample estimates. The Hadamard matrix
was used to derive replicate factors to apply to individual full sample weights in creating
replicate weights.
The goal of this example is to estimate the total number of renter-occupied housing units in
Queens for 2017 and its corresponding estimate of variance.
For example, we have 1,810 completed interviews that are renter-occupied housing units in
Queens. Table 6.1 displays the first four and last one renter-occupied sample units in Queens.
The result show below are the sample cases in Queens with responses to tenure status question as
renters.
Table 6.1: Example of Estimating Variances with Replication
Replicate Weights
Sample
HU

Full Sample
Weight

Replicate
1

Replicate
2

Replicate
3

Replicate
80

1

250.430

250.430

234.225

75.769

…

272.506

2

241.448

241.448

224.532

254.145

…

263.398

3

240.695

240.695

416.378

225.885

….

74.076

4

178.260

178.260

303.175

184.240

…

52.920

…

…

…

…

…

1810

11.598

11.598

3.566

10.865

…
…

11.525

In NYCHVS, the full sample estimate and the full sample weight are referred to as the replicate
estimate 0 and replicate weight 0, respectively.
Step 1: Calculate the full sample weighted survey estimate.
The statistic of interest is the total number of renter-occupied housing units in Queens for 2017.
Add the full sample weights of the sample cases that meet your criteria of interest. Therefore,
the total number of renter-occupied housing units in Queens is calculated as follows:
Full-Sample Renter-Occupied in Queens Estimate:
29

Nˆ = 250.430 + 241.448 + … + 11.598 = 439,257.02

Step 2: Calculate the weighted survey estimate for each of the replicate samples.
The replicate survey estimates are as follows:
Replicate 1 Renter-Occupied Estimate
Replicate 2 Renter-Occupied Estimate

Nˆ r 1 = 250.430 + 241.448 + … + 11.598 = 439,257.02
Nˆ = 234.225 + 224.532 + … + 3.566 = 440,785.37

Replicate 3 Renter-Occupied Estimate

Nˆ r 3 = 75.769 + 254.145 + … + 10.865 = 435,992.59

r 2

.
.
.

.
.
.

.
.
.

̂𝑟=80= 272.506 + 263.398 + … + 11.525 = 436,801.68
Replicate 80 Renter-Occupied Estimate 𝑁
̂𝑟 in the formula below to calculate the variance estimate for
Step 3: Use the replicate estimates 𝑁
the total renter-occupied HUs in Queens.
80

4
̂) =
̂𝑟 − 𝑁
̂0 )2
𝑣̂(𝑁
∑(𝑁
80
𝑟=1

= 0.05 × [(439,257.02 − 439,257.02)2 + (440,785.37 − 439,257.02)2
+ (435,992.59 − 439,257.02)2 + ⋯ + (436,801.68 − 439,257.02)2 ]
= 0.05 × [0 + 2,335,853.72 + 1,0656,503.22 + ⋯ + 6,028,694.52]
= 29,362,077.47
The estimate of the variance of total renter-occupied HUs in Queens is vˆ( Nˆ ) =29,362,077.47.
The survey estimate for total renter-occupied population in Queens is 439,257.02 housing units.
This survey estimate has an estimated variance of 29,362,077.47, or a standard error of 5,418.68
housing units.

7. References
Fay, R. E. and Train, G. F. (1995). Aspects and Survey and Model-based Postcensal Estimation
of Income and Poverty Characteristics for States and Counties. Proceeding of the Sections on
Government Statistics, American Statistical Association, 154-159.
Ash, S. E. (2014) Using Successive Difference Replication for Estimating Variances. Survey
Methodology, Statistics Canada, Catalogue no.12-001-X Business Survey Method Division, Vol.
40, No.1, pp.47-59.
30

Lohr, S.L. (2007). Recent developments in multiple frame surveys. Proceedings of the Survey
Research Methods Section, American Statistical Association, 3257-3264. Accessed online at
http://www.amstat.org/sections/srms/Proceedings/ on September 1, 2015.
Lohr, S. (2010). “Dual Frame Surveys: Recent Developments and Challenges,” paper presented
at the Scientific Meeting of the 45th Italian Statistical Society, Padua, Italy, June 16-18.
U.S. Census Bureau (2014). Overview. https://www2.census.gov/programssurveys/nychvs/about/overview/overview-2014. Date retrieved April 23, 2018.
U.S. Census Bureau (2017a). Methodology for United States Population Estimates: Vintage
2017. https://www2.census.gov/programs-surveys/popest/technicaldocumentation/methodology/2010-2017/2017-natstcopr-meth.pdf Date retrieved April 23, 2018.
U.S. Census Bureau (2017b). Overview. https://www2.census.gov/programssurveys/nychvs/about/overview/overview-2017. Date retrieved April 23, 2018.

31

Appendix A. Example of Ratio Adjustments
This appendix provides one hypothetical example that demonstrates how the sample weights
were adjusted so that they were consistent with a set of control totals. The example is a ratio
adjustment.
For this example, assume weights were calculated for a sample and the weights included all
weighting adjustments up to a nonresponse adjustment. With these weights, totals by two
categories of tenure status (owner or renter) and two categories of type of construction (old or
new) were created. Table A1 summarizes the estimated totals resulting from the hypothetical
sample and weights.

Table A1: Estimated Totals
Owners

Renters

Total

New

110

91

201

Old

97

107

204

Total

207

198

405

Suppose the control totals were as shown in table A2.

Table A2: Example 1 Control Totals
Owners

Renters

Total

New

115

105

220

Old

95

105

200

Total

210

210

420

The control totals of table A2 are used to improve the weights by making the estimates from the
weights consistent with the control totals. Table A3 shows the Ratio Adjustment Factor (RAF)
that will make the estimated totals consistent with the control totals.

Table A3: Example 1 Ratio Adjustment Factors
Owners

Renters

New

115/110 = 1.0455

105/91 = 1.1583

Old

95/97 = 0.9794

105/107 = 0.9813

If the factors from Table A3 are applied to the weights of the sample units, then the estimates
from the revised weights will be consistent with the totals of table A2.
A-1

Note that ratio-adjusted weights for the combination of owners and new construction is the
product of the weight before the RAF, that is,
Ratio-adjusted weight = original weight ´ 1.0455 .
The ratio-adjusted weights for the other three cells are defined similarly.

A-2

Appendix B: List of Variables Imputed for 2017 NYCHVS
Table B1: List of Variables Imputed for Occupied Units
Occupied Units
Item Name
Variable Name
Additional Source(s) of Heat
SC187
Any Buildings with Broken or Boarded-Up
SC24
Windows (Observation)
Broken Plaster or Peeling Paint on Ceiling or
SC192
Inside Walls
Broken Plaster or Peeling Paint on Ceiling or
SC193
Inside Walls Larger than 8 1/2 x 11
Borough
BORO
Combined Gas/Electricity Cost
UF14
Complete Kitchen Facilities
UF83
Complete Plumbing Facilities
UF81
Condition of Building (Observation)
SC23
Condition of External Walls
UF1_1, UF1_3 - UF1_6
Condition of Floors
UF1_17, UF1_19 UF1_22
Condition of Stairways (Exterior and Interior)
UF1_12 - UF1-16,
UF1_35
Condition of Building Recode
REC21
Condition of Windows
UF1_7 - UF1_11
Condo/Coop Before Acquisition
SC121
Condo/Coop Conversion Done Through a
SC118
Non-Eviction Plan
Condo/Coop Status
SC114
Control Status Recode
UF_CSR
Cracks or Holes in Interior Walls or Ceiling
SC190
Down Payment
UF5
Electricity Paid Separately
SC159
Electricity Monthly Cost
UF12
Exterminator Service
SC189
Federal, State, or Local Government Payments for
SC184, SC541-SC544
Rent
Fire and Liability Insurance Paid Separately
SC141
First Occupants of Unit
SC54
Floor the Unit is On
UF50
Functioning Air Conditioning
SC197
Gas Paid Separately
SC161
Gas Monthly Cost
UF13
General Health of Respondent
SC574
Heating Equipment Breakdown
SC185
Holes in Floors
SC191
Household Below Specified Income Level Recode
REC39
Household Composition Recode
REC1

B-1

Imputed?
No
No
No
No
No
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
No
Yes
Yes
No
No
No
No
No
No
Yes
Yes
No
No
No
Yes
Yes

Occupied Units
Item Name
Variable Name
Household Income from Farm or Nonfarm
UF35
Business, Proprietorship, or Partnership Recode
Household Income from Retirement, Survivor, or
UF38
Disability Pensions Recode
Household Income from Social Security or
UF37
Railroad Retirement Payments Recode
Household Income from SSI, TANF/Family
UF39
Assistance, Safety Net, or Other Public Assistance
or Public Welfare Payments, Including Shelter
Allowance Recode
Household Income from VA Payments,
UF40
Unemployment Compensation, Child Support,
Alimony, or Other Source of Income Recode
Household Income from Wages, Salaries,
UF34
Commissions, Bonuses, or Tips Recode
Household Member Under Age of 6
UF75
Household Member Under Age of 18
REC7
Householder of Spanish/Hispanic Origin
HHR5
Householder Moved to the United States as
SC560
Immigrant
Householder's Age Recode
UF43
Householder's Race
UF61
Householder's Sex
HHR2
Kitchen Facilities Functioning
SC157
Length of Lease
SC181
Householder Lived in Unit at Time of
SC117
Coop/Condo Conversion
Medical Device in Home
SC198
Monthly Contract Rent
UF17
Monthly Contract Rent as a Percent of Household
UF29
Income Recode
Monthly Contract Rent per Room Recode
UF27
Monthly Gross Rent
UF26
Monthly Gross Rent as a Percent of Household
UF30
Income Recode
Monthly Gross Rent per Room Recode
UF28
Monthly Owner Cost Recode
UF105
Mortgage Interest Rate (Current)
UF7A
Mortgage Origination Year
UF68
Mortgage Status
SC127
Most Recent Place Householder Lived for 6
UF79
Months or More
Number of 1987 Maintenance Deficiencies
REC54
Recode
Number of 2017 Maintenance Deficiencies
REC53
Recode
Number of Bedrooms
UF78

B-2

Imputed?
Yes
Yes
Yes
Yes

Yes

Yes
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
No
No
No
No
Yes

Occupied Units
Item Name
Variable Name
Number of Cockroaches
SC571
Number of Heating Equipment Breakdowns
SC186
Number of Persons from Temporary Residence
UF2A_1
Number of Persons per Room Recode
CPPR
Number of Persons Recode
UF73
Number of Rooms
UF77
Number of Units in Building
UF47
Number of Stories in Building
UF11
Occupancy Status Before Acquisition
SC120
Other Fuels Paid Separately
SC166
Other Fuels Annual Cost
UF16
Out of Pocket Rent
UF17A
Owner Lives in Building
SC147
Passenger Elevator in Building
SC149
Place of Householder's Birth
SC111
Place of Householder's Father's Birth
SC112
Place of Householder's Mother's Birth
SC113
Postponement of Health Care for Financial
SC647-SC651
Reasons
Presence of Mice and Rats
SC188
Presence of Nonrelatives Recode
UF46
Receipt of Public Assistance or Welfare Payments SC548-SC551, SC175,
SC199
Race and Ethnicity of Householder Recode
UF69
Race Recode 1 (Householder)
UF60
Real Estate Taxes Paid Separately
SC144
Respondent Line Number
UF71
Respondent Rating of Residential Structures in
SC196
Neighborhood
Respondent Opinions of their Housing Unit's
SC168, SC169, SC183
Affordability
Senior Citizen Carrying Charge Increase
SC184
Exemption (SCRIE)
Service Interruptions for Financial Reasons
SC131, SC132, SC136,
SC137, SC138
Sidewalk to Elevator Without Using Steps or
SC173
Stairs
Sidewalk to Unit Without Using Steps or Stairs
SC171
Structure Class Recode
UF74
Sub-borough Area
CD
Telephone (Landline) in Apartment (House)
SC575
Tenure (1) Owner/Renter
SC115
Tenure (2) Cash Rent/Rent Free
SC116
Toilet Breakdowns
UF82
Total Household Income Recode
UF42
Type of Heating Fuel
SC158
Type of Schedule Code
UF76

B-3

Imputed?
No
No
No
Yes
No
Yes
Yes
Yes
No
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
Yes
No

Occupied Units
Variable Name
Value
UF6
Water and Sewer Paid Separately
SC164
Water and Sewer Annual Cost
UF15
Water Leakage Inside Apartment
SC194
Wheelchair Accessibility
SC36, SC37, SC38
Year Built Recode
UF23
Year Householder Moved Into Unit
UF66
Year Householder Moved to the United States
UF53
Year Householder Moved to New York City
UF54
Item Name

B-4

Imputed?
Yes
Yes
Yes
No
No
Yes
Yes
No
No

Table B2: List of Variables Imputed for Persons
Persons
Variable Name
Age Recode
UF43
Average Hours Worked per Week in 2016
UF96
Borough
BORO
Check Item G
CHK_G
Current Enrollment in Job Training/Education
ITEM50A
Educational Attainment
EDUCTN
Hours Worked Last Week
UF95
Income from Own Farm or Nonfarm Business,
UF18A
Proprietorship, or Partnership
Income from Interest, Dividends, Net Rental or
UF18B
Royalty Income, or Income from Estates and
Trusts
Income from Retirement, Survivor, or Disability
UF18E
Pensions
Income from Social Security or Railroad
UF18C
Retirement Payments
Income from SSI, TANF/Family Assistance,
UF18D
Safety Net, or Other Public Assistance or Public
Welfare Payments, Including Shelter Allowance
Item Name

Income from VA Payments, Unemployment
Compensation, Child Support, Alimony, or
Other Source of Income
Income from Wages, Salaries, Commissions,
Bonuses, or Tips
Labor Force Status Recode
Last Time Worked
Looking for Work
Major Industry Type
Number of Weeks Worked in 2016
Person from Temporary Residence
Person Number of 1st Parent
Person Number of 2nd Parent
Person Number of Spouse/Partner
Race
Race and Ethnicity of Householder Recode
Race Recode 1
Relationship
Sex
Spanish/Hispanic Origin
Sub-borough Area
Temporarily Absent or on Layoff from a Job
Last Week
Total Person Income Recode
Type of Worker
Workers' Industry Code

B-5

Imputed?
Yes
Yes
No
Yes
No
Yes
Yes
Yes
Yes

Yes
Yes
Yes

UF18F

Yes

UF18

Yes

UF59
ITEM44
ITEM42
ITEM45C
ITEM48A
UF3
UF87
UF88
UF86
UF62
UF70
UF60
UF92
SEX
HSPANIC

Yes
Yes
Yes
Yes
Yes
No
No
No
No
Yes
Yes
Yes
No
Yes
Yes

CD
ITEM41

No
Yes

UF41
UF90
UF94

Yes
Yes
Yes

Item Name
Worked Last Week
Workers' Occupation Code
Year Non-Householder Moved Into Unit

Persons
Variable Name
ITEM40A
UF93
UF55

B-6

Imputed?
Yes
Yes
No

Table B3: List of Variables Imputed for Vacant Units
Item Name
Any Buildings with Broken or Boarded-Up
Windows (Observation)
Borough
Complete Kitchen Facilities
Complete Plumbing Facilities
Condition of Building (Observation)
Condition of Building Recode
Condition of External Walls
Condition of Floors

Vacant Units
Variable Name
SC24

Condition of Stairways (Exterior and Interior)
Condition of Windows
Condo/Coop Status
Condo/Coop Status Before Vacancy
Control Status Recode
Duration of Vacancy
First Occupancy
Floor the Unit is On
Monthly Asking Rent
Monthly Asking Rent per Room Recode
Number of Bedrooms
Number of Rooms
Number of Units in Building
Owner in Building
Passenger Elevator in Building
Reason Unit Not Available for Rent or for Sale
Sidewalk to Elevator without Using Steps or
Stairs
Sidewalk to Unit without Using Steps or Stairs
Status of Vacant Unit
Status Prior to Vacancy
Stories in Building
Structure Class Recode
Sub-borough Area
Type of Heating Fuel
Type of Schedule
Vacant Unit Respondent
Wheelchair Accessibility
Year Built Recode

Imputed?
No

BORO
UF84
UF91
SC23
REC21
UF1_1, UF1_3 - UF1_6
UF1_17, UF1_19 UF1_22
UF1_12 - UF1-16,
UF1_35
UF1_7 - UF1_11
SC530
SC533
UF_CSR
SC531
SC518
UF51
UF31
UF32
UF78
UF77
UF47
SC520
SC522
UF80
SC553

No
Yes
Yes
No
No
No
No

SC555
SC534
SC532
UF33
UF74
CD
SC529
UF76
SC30
SC36, SC37, SC38
UF23

No
No
No
Yes
No
No
Yes
No
No
No
No

B-7

No
No
No
No
No
No
No
No
Yes
Yes
Yes
Yes
Yes
No
No
No
No

Appendix C: Housing Unit Characteristics Associated With GVF Parameters
For characteristics and subgroups matching to this table (Table C1), use the second of the three
sets of parameters from the housing unit GVF parameters (Table 5.1a).
Table C1: Housing Unit Characteristics Associated with the Second of Three Sets of Parameters
on Table 5.1a
Characteristics
Applicable Subgroups
Race and Ethnicity of Householder
(White, non-Hispanic and African American, nonHispanic)

Total Housing Units

Borough Totals

Renter Occupied (Stabilized, Mitchell
Lama, Public Housing) and
Owner Occupied (Condominiums and
Total Cooperatives)

Sub-borough of Staten Island Totals

Total Housing Units, Total Occupied
Housing Units, Total Rental Housing
Units and Total Occupied Rental
Housing Units

Contract Rent < $300

Total Housing Units and Total
Occupied Housing Units

Wheel Chair Accessibility

All subgroups except

Floor Unit is on (except basement)

Renter Occupied - Controlled and

Access from Sidewalk to Elevator/Unit without
using Stairs

Owner Occupied - Conventional

Households Not Receiving Part of Monthly Rent
from Government Programs
Condition of Building External Walls, Windows,
Stairways, and Floors of Building
Number of Building Condition Problems 1-4

C-1

Total Occupied and Total Renter
Occupied

For characteristics and subgroups matching to this table (Table C2), use the third of the three sets
of parameters from the housing unit GVF parameters (Table 5.1a).
Table C2: Housing Unit Characteristics Associated with the Third of Three Sets of Parameters
on Table 5.1a
Characteristics
Applicable Subgroups
Sub-borough Totals (All Boroughs Except Staten
Island)

Total Housing Units, Total Occupied
Housing Units, Total Rental Housing
Units and Total Occupied Rental
Housing Units

Structure Classification - Multiple dwelling units

Total Housing Units and
Total Occupied Housing Units

Structure Classification - One or 2 family house

Total Housing Units

Rent Control Status

Total Rental Housing Units and Total
Occupied Rental Housing Units

Year Building Built

Total Occupied and Total Renter
Occupied

Number of Stories in Building
Number of Units in Building
Presence of Owner in Building
Elevator in Building with 2 or more stories
State/City Assisted Cooperatives

Total Owner Housing Units and Total
Occupied Owner Housing Units

Private Cooperatives
Private Condominiums

C-2