Docstoc

schooling

Document Sample
schooling Powered By Docstoc
					The U.S. Census Bureau’s Small Area Estimates Program:
       Current Approaches and Future Prospects

                            By

                     William Schooling
            Chief, Population Estimates Branch,
                    Population Division
                    U.S. Census Bureau



                        Prepared for

         The Census of Population: 2000 and Beyond
             Manchester, U.K. 22-23 June 2000
II     INTRODUCTION

        The Population Estimates Program at the U.S. Census Bureau is located within the
Estimates and Projections area of the Population Division. We have five branches in the
Estimates and Projection section. These branches are responsible for a very full data production
schedule, not to mention research and decennial evaluation work. Our national estimates are
prepared in the Projections Branch. The subnational estimates are produced in the Population
Estimates Branch. However, much of this work requires cooperation among all of the branches.
Since the focus of this paper is on small area estimates, I should also mention that we work with
a governor appointed demographer in each state to assist in the production of estimates. The
Federal State Cooperative Program for Population Estimates (FSCPE) provides us with a source
of local demographic insight.

II     BRIEF OVERVIEW OF U.S. NATIONAL ESTIMATES

       Before we get to the US small area estimates program, I thought I should do a very, very
quick overview of the not-so-small area estimates program. While we separately prepare the
population estimates for states, counties, and subcounty areas; in the end, the sum of these
estimates must equal the independently prepared national estimate.

       The program to develop the national population estimates is benchmarked from the 1990
census. None of the Bureau’s estimates incorporate an adjustment for net census undercount. In
preparing the national estimates, we employ the basic demographic accounting–cohort
component approach. Starting with 1990 census as the base, we add annual births, subtract
annual deaths and add our estimates of net international migration.

        The international net migration estimates are based on various administrative sources and
estimates. The Immigration and Naturalization Services (INS) supplies data on legal
immigration. The Office of Refugee Resettlement provides additional information. Working with
data from INS and the decennial censuses, the Bureau prepares an estimate of annual
undocumented immigration and emigration. Since we prepare estimates of the resident
population of the U.S., the last part of the accounting equation estimates the net movement of
Armed Forces, civilian Federal employees, and dependents of both, which we obtain from the
Department of Defense and the Office of Personnel Management.

III    COUNTY/STATE METHODOLOGY, SHORT VERSION

        We use a bottom-up approach to prepare our state estimates. That is, our state estimates
are simply the sum of the county population estimates within any given state. This has not always
been the case and it may change at some point in the future, but for the 90's, that has been our
approach.
        I will provide slightly more detail on our county estimates program because it is relevant
to our small area estimates. We start with our benchmark, the 1990 census. We use a component
method to prepare the annual July 1 estimates. The current U.S. county method is called the Tax
Return method, formerly the Administrative Records Method. Because the state estimates for
July 1 of each year must be completed before the end of the calendar year, we essentially finish
our county estimates (with only minor adjustments) also by the end of the year.

        Because some of the necessary data are not available by the time we must produce the
estimates, some data components are extrapolated. For example, our births and deaths are usually
not available at the county or state level for the last six months of the estimates’ period (January
1 to June 30 of the current year). The method of extrapolation varies according to the
circumstances. Because we must extrapolate some of the components to the current July 1 date,
some would say our estimates are really a combination of estimate and projection techniques.
However, these extrapolations play only a very small part in our overall scheme.

        I will briefly discuss our county estimates, but there are more details in the appendices of
the paper. Table 1 describes the methodology for estimating the annual July 1 populations under
age 65. The process separates out the group quarters population and the population 65 years old
and older. Then it becomes a basic component method of births, deaths, and migration to
complete the under 65 years estimate. The most important dimension of change, normally, is the
calculation of domestic migration. We must develop a rate of net migration. To develop this rate,
we rely upon a match of the annual tax return data, which we receive from our tax office–the
Internal Revenue Service (IRS). By matching on returns using social security numbers, we
develop an estimated rate for each county’s net domestic migration.

        Table 2 provides an explanation of the county estimates for those 65 and over. It is not
very different from the under 65, except the primary data source for migration is Medicare
(health insurance program) data that is more complete than tax records for those in the middle to
upper age brackets (65 and over). Death statistics are readily available from State and National
sources, but births into the cohort (those turning 65 years of age in the year prior to the estimate
date) have to be estimated.

         Table 3 shows the development of the final estimate. The estimated county populations
for those under 65 are raked to the national estimate for the same age group. So the sum of each
county will match our national estimate. The same method, with a different rake factor is used
for those 65 and over. The sum of the two age groups, after their rakes, is the final estimate, at
least for the first phase. We review the output, as well as the change components, and we also ask
state members of the Federal State Cooperative Program for Population Estimates to review the
same data for counties in their state. We review their comments and take the comments of our
own analysts and make changes we think are appropriate. The counties, which change, override
the first estimate and because we rake to the national totals, all other counties are effected by a
second rake, but these secondary changes are small. Remember, the state estimates are sums of
the county estimates after the rake is performed, so state estimates will also add to the nation.
Table 1.

Derivation of 1998 Under 65 Population Estimate for:

BASE POPULATION
 1. Base Total Population for Current Year             112612     See text for detailed source

 2.   Base Group Quarters Population
      Under Age 65 for Previous Year                   12191      See text for detailed source

 3.   Base Total Population Age 65 and Over            10236      See text for detailed source
      for Previous Year

 4.   Household Base Population Under Age 65           89538      (4) = (1) - (2) - (3) - [(.06320828)*3]
      for Current Year

COMPONENTS OF CHANGE FOR THE HOUSEHOLD POPULATION UNDER AGE 65

 5.   Estimated Resident Births:
      7/1/97 to 6/30/98                                1214           See text for detailed source

 6.   Estimated Resident Deaths to the
      Household Population Under Age 65                165            See text for detailed source

 7.   Estimated Foreign Immigration Under Age 65:      163            See text for detailed source
      7/1/97 to 6/30/98

 8.   Estimated Migration Base for Under Age 65        90144          (8) = (4) + {0.5 x [(5) - (6) + (7)]}

 9.   Estimated Migration Rate Under Age 65            -0.00564       See text for detailed source

10.   Estimated Net Migration Under Age 65             -508           (10) = (8) x (9)

11.   Estimated Household Population                   90242          (11) = (4) + (5) - (6) + (7) + (10)
      Under Age 65

12.   Estimated Group Quarters Population              14949          See text for detailed source
      for Current Year

13.   Estimated total population under 65              105191         (13) = (11) + (12)

                     Source:       Population Estimates Branch
                            U.S. Bureau of the Census
Table 2.

Derivation of 1998 65 and Over Population Estimate for:


BASE POPULATIONS
 1. Base Total Population Age 65 and Over          10236         See text for detailed source
     for Previous Year

 2.   Base Group Quarters Population
      Aged 65 and Over for Current Year            391           See text for detailed source

 3.   Estimated Population Turning 65
      for Current Year                             647           See text for detailed source

 4.   Household Base Population Age 65 and Over    10492         (4) = (1) - (2) + (3)
      for Current Year


COMPONENTS OF CHANGE FOR THE HOUSEHOLD POPULATION       AGED 65 AND OVER

 5.   Estimated Resident Deaths to the
      Household Population Aged 65 and Over        509           See text for detailed source

 6.   Estimated Foreign Immigration:               -1            See text for detailed source
      7/1/97 to 6/30/98

 7.   Estimated Migration Base                     10237         (7) = (4) + 0.5 x [(6) - (5)]

 8.   Estimated Migration Rate                     0.003446      See text for detailed source

 9.   Estimated Net Migration                      35            (9) = (7) x (8)

10.   Estimated Household Population               10017         (10) = (4) - (5) + (6) + (9)
      Aged 65 and Over

11.   Estimated Group Quarters Population          391           See text for detailed source
      for Current Year

12.   Estimated Total Population Age 65 and Over 10408           (12) = (11) + (10)

                   Source:       Population Estimates Branch
                         U.S. Bureau of the Census


Table 3.

Final estimate for:


1.   Estimated Total Population Under Age 65           105191     Line 13 from Table 1

2.   Rake Factor for Total Population Under Age 65     1.000106   See text for explanation

3.   Final Estimate of Population Under Age 65         105202     (3) = (1) X (2)

4.   Estimated Total Population                        10408      Line 12 from Table 2
     Age 65 and Over

5.   Rake Factor for Total Population                  0.99805    See text for explanation
     Age 65 and Over

6.   Final Estimate of Population                      10388      (6) = (4) X (5)
     Age 65 and Over

7.   Final Estimate for Total Population               115590     (7) = (3) + (6)



                   Source:     Population Estimates Branch
                         U.S. Bureau of the Census
IV     CURRENT SUBCOUNTY ESTIMATES PROGRAM


        We often use the term subcounty to talk about small area estimates. Of course we can
only blame some of you for that. When Europeans arrived in the New World they brought their
own ideas about governing bodies with them. This is not considered a federal issue, so subcounty
geography varies considerably among states. In some states we have only counties and cities,
while in others there are entities such as townships, towns, villages and boroughs. The
boundaries of these various governing bodies may overlap, which makes things even more
interesting. We have pieces of towns crossing townships crossing county boundaries. Therefore,
we prepare estimates for all of these pieces in order to be able to summarize up to whole
governmental units (e.g., a city). It adds up to about 78,000 governmental units and pieces which
are less than an entire county. These 78,000 pieces represent our small area estimates universe.

        County population estimates are still a key in developing small area estimates. We use a
housing unit method to distribute county population totals to each small area in our estimates
universe. In Table 4, you have what appears to be a basic method and this is the core of the
process. However, exceptions, requiring special automated procedures, are not that unusual and
regrettably not all data sets are as clean as we would like. Additionally, some of the variables
used in constructing the housing unit estimates entail allocation routines. For example, while we
generally rely upon data on building permits to estimate residential construction, we have many
cities which do not require or do not report building permits. Since construction is a key
component in the method, we have to be able to supplement our data with other information for
these areas. Similarly, housing unit loss would often be estimated using housing demolition data.
However, demolition data are often suspect. Although the Census Bureau no longer collects these
data, we still have to estimate it. Other factors used in the model may vary for different parts of
the nation. One such factor is the lag time between when a building permit is issued and when
the unit is ready to be occupied.

       Lines six and seven of Table 4 show that we are using each locality’s 1990 occupancy
and persons per household rate. This is generally true, but if computer edits show that there has
been a change outside of established bounds, adjustments are automatically made. From survey
data we do track regional change, however, we do not have firm local data on these changes. We
are conducting research which we hope will improve estimates of occupancy and persons per
household change throughout the decade for all incorporated areas and counties.

       Group quarters (GQ) estimates for small areas are also based on a change method. The
state members of the Federal State Cooperative Program for Population Estimates ( FSCPE)
provide information annually on the population of group quarters in their states. However,
because they are unable to gather data on all group quarters identified in the last decennial, the
updated information applies to only a subset of all group quarters within the state. Often there are
substantial differences between what the FSCPE find and what the census found. To
accommodate for this, we add the change in the population for the units reported to the census
GQ data to arrive at a new GQ estimate. These estimates are also updated annually, so change
between 2000 and 2001 will be added to the census, change between 2001 and 2002 will modify
the 2001 estimate.
        The county rake factor is developed by summing the total of all small area estimates
within a given county and then using the appropriate proportion to distribute the county
population to each of these areas. An exception to the rake occurs because of selected
administrative changes. One of these exceptions is due to our policy of allowing local challenges
to our population estimates. If a local government questions the population estimate, we provide
them with a complete package of the data we used, with instructions on alternate methods or data
sources they can provide. If the local government submits either alternate data or can provide
data and employ one of the accepted alternate methods, we review the returned materials and
often accept the new estimate. In this case, we would exclude the area from the rake for that
county.

Details of our variation of the housing unit method can be found in Appendix B.
                                                    TABLE 4

Derivation of a Current Year Population Estimate for a Hypothetical Local Government:

                                                                Value      Derivation or Source
------------------------------------------------------------------------
1. Housing Units in 1990 .......                                14,240     From Decennial Census

2. Estimated Residential Construction: 4/1/90 to
   Estimate Year...............                                2,010       See text for detailed source

3. Estimated New Mobile Home Placements:
   4/1/90 to Estimate Year...............                      88          See text for detailed source

4. Estimated Housing Unit Loss
   4/1/90 to Estimate Year ....                                28          See text for detailed source

5. Housing Units in Estimate Year .................            16,310      (5) = (1) + (2) + (3) - (4)

6. 1990 Occupancy Rate ...........                             .978        From Decennial Census

7. 1990 Persons per Household...                               2.89        From Decennial Census

8. Uncontrolled Household Population Estimate                   46,099     (8) = (5) * (6) * (7)

9. Uncontrolled County Household
   Population Estimate ........                                759,159     See text for detailed source

10. Published County Estimate...                               744,415     See text for detailed source

11. County Rake Factor..........                               .98057      (11) = (10)/(9)

12. Controlled Household Population Estimate                   45,203      (12) = (8)*(11)

13. Group Quarters Population Estimate..........               1,685       See text for detailed source

14. Final estimate..............                               46,888      (14) = (12) + (13)


                     SOURCE: Population Estimates Branch, U.S. Census Bureau
V      RESEARCH AND POTENTIAL CHANGES TO CURRENT METHODOLOGIES

        What drives change in population? Any demographer can answer that question. At any
time in the history of humankind has it ever been anything other than births, deaths, and
migration? There have been individuals who have been interested in demographic data for
thousands of years and we have had individuals studying the science of demography for hundreds
of years. A professor of mine once remarked that he didn’t believe there would be any
demographic change if you looked 300 years ahead. I disagree with him, partly because we
would be out of a job by then, if he is right. Whether or not this professor is right or wrong, it is
true that the formula for population change does not change. Then why does the science of
demography change? I believe that there are a couple of answers to that question. One is that our
techniques and tools become more refined, we hope. The other driving force, in my opinion, are
the users of our data. Their axiom always seems to be more, smaller, better, and faster, in other
words we are given expanding data requirements and we have to find the means to answer their
demands.

       An example of more are the growing requirements we face for added racial and ethnic
data. With recent changes in racial classifications--to all possible combinations of white, Black,
American Indian or Alaska Native, Asian, and Native Hawaiian--we will be forced to produce
population estimates for up to 32 possible racial categories. Coming up with the appropriate
component data to move forward after Census 2000 will be a challenge. We need to look for new
data sources and introduce additional modeling efforts.

        Smaller means lower level geographic units. We anticipate (fear) the day that we will
need to prepare annual block level or even lower level estimates. The demand is usually
generated by data users who want these building block estimates to sum to areas that are of
greater interest to them, but are outside of the estimates universe. School districts or
congressional districts are two examples. We even discuss individual housing units as a potential
geographic level for estimates to use for aggregation purposes.

        Faster or at least more often is a common theme among our users and no doubt yours as
well. There is always demand to release the estimates closer to the official date of the estimate.
Over the years, the Census Bureau has not only added to the universe of areas for which
estimates are prepared, but we have started producing estimates on a more regular basis. The
prime example is our small area estimates program where we have moved from a once-in-awhile
release schedule to once a year.

        Better quality, well who doesn’t want better quality? The Estimates Branch has instituted
numerous quality improvement measures, including automated routines at most steps in the
process. Additionally, we are using more tools, such as the loss function to detect outliers, so that
we can more closely examine potential problematic areas. Automated quality steps and additional
tools can help improve the accuracy of the estimates. However, we thought we could do more.


       The Population Estimates Branch, the Administrative Records and Methodological
Research Branch, and other branches within the Population Division, are pursuing a fairly
extensive set of research projects. Additionally other divisions in the Census Bureau, especially
the Program Research and Evaluation Division (much of their work involves a variety of
administrative records) are working on their own projects and we plan to incorporate as much of
their work as we can. These projects are designed to take advantage of new data sets, such as the
American Community Survey, new administrative data sources, and geographic information
system products; to reevaluate existing data from sources we do not use; to develop better
methods of evaluating estimates so that we are confident that we are using the best available
methodology, and the best available data sets.

         The Census Bureau held a very interesting conference in 1999, where speakers talked
about their own methods or research. A few of the speakers are here today and I will be interested
in talking with them to see how their research is progressing. Some speakers were from academic
institutions, some were local demographic consultants, and many others were the FSCPE
representatives. It seems that we face similar problems, but find somewhat different approaches
to move toward solutions of our challenges. One of the purposes of that conference was to spur
creative juices for research in preparation for estimates in the post 2000 period. I am sure this
conference will provide additional insight.

         In the U.S., it does seem to be rare to find a completely new method to develop estimates,
especially one that can be applied throughout the nation, whether that means all counties, all
states, or all small area estimates. The Census Bureau started a quest to identify what methods
various agencies were using, whether for just a few areas or with national applicability. Even
though we use state demographers in the FSCPE to provide some data and to review components
and estimates, perhaps more involvement will increase the quality of the estimates.

        Essentially we have challenged the state demographers to prepare a set of 2000 estimates
to be evaluated, along with ours, and compared to Census 2000. This competition is open to
anyone, although we expect the FSCPE members to be the primary participants. We asked the
FSCPE to name a panel of members to work with the Bureau on designing a level playing field
of competition in preparing estimates for evaluation. The FSCPE panel will meet with us in
August to be sure that the evaluation techniques are fair to everyone. In comparing various sets of
estimates to Census 2000, we expect to answer at least a few questions. Will unique state data
equal better estimates? Would an average, of some type, of their methods and our methods
improve the quality? We may find that a few methods do better than the Bureau produced
estimates. We expect that these “winners” may only be attempting to apply their method to a
subset of areas, such as all counties within one state. Because of the complexity of
implementation, we do not look forward to the introduction of new methods (or simply the
estimates produced by other methods) that do not apply uniformly to all small areas or all
counties, but it is our decision that if this improves the quality of estimates, we will do just that.
Source:

Most of the data tables and appendices presented in this paper can be found at the Census Bureau
website: www.census.gov Look for “People” then select “Estimates.”

Acknowledgments:

I greatly appreciate the review, comments and contributions of Gregory Harper, Sam Davis, and
Signe Wetrogan of the Population Division of the Census Bureau.
Appendix A–

Explanations Table 1: Estimates of the Under Age 65 Population

Base Populations

TABLE 1- Line 1
Base total population--The base population for the estimate of the population under age 65 is the
revised county estimate for the prior estimate year. Each year, the population estimate represents
population change from the prior year. The only year in which this is not true is in the year of the
decennial census. In the decennial year, an estimate is prepared that represents population
change between the census date and July 1 of that year. For example, the last decennial census
was collected as of April 1, 1990. A population estimate was prepared that represented
population change between April 1, 1990 and July 1, 1990. For official population estimates, the
April 1, 1990 population is not adjusted for census undercount.

   a. Census errata--Geographic and enumeration errors in the 1990 census for some areas have
been identified and subsequent corrections made. These errors may have an impact on
population estimates in some counties. Corrections to the 1990 census were first incorporated
into the estimates series with the release of estimates for July 1, 1994.

   b. Census modifications--The 1990 census population counts were modified to handle two
areas problematic for estimates development: (1) We modified the census results so that persons
responding "other" to the 1990 census race question were reclassified into one of the four major
race classifications:

       White, Black, Asian/Pacific Islander (API) or American Indian/Alaska Native (AIAN);
       (2) The population age distribution was also modified, owing to a misreporting problem.
       For more information about these modifications, call the Census Bureau at
       (301) 457-2422 and ask for a copy of the technical documentation developed for the
       modified age/race/sex file (MARS).

    c. Boundary changes--Boundary changes include new incorporations, disincorporations,
annexations, deannexations, mergers and consolidations. In general, county population
estimates are consistent with boundaries current as of the estimate date. Boundary changes
occurring after January 1, 1990 were incorporated into the population estimates base beginning
with estimates for July 1, 1994.

TABLE 1--Line 2
Base group quarters population under age 65--This component is primarily a combination of
military personnel living in barracks, college students living in dormitories and persons residing
in institutions. Inmates of correctional facilities, persons in health care facilities and persons in
Job Corps Centers are also included in this category. We have excluded persons aged 65 and
over residing in nursing homes and other facilities from this category since they are implicitly
included in the estimate of the 65+ population. The base group quarters population for the
current estimate year is the revised group quarters population from the prior estimate year. In the
first estimate year following the decennial census, the base group quarters population is the group
quarters population as enumerated in the decennial census.

TABLE 1--Line 3
Base total population aged 65 years and over--This component is the revised estimate of the
population aged 65 years and older from the prior estimate year.

TABLE 1--Line 4
Household base population under age 65--After we subtract the group quarters populations (Line
2) and the population aged 65 and over (Line 3) from the base population (Line 1) the remainder
is what we call the under 65 household population. The household population under age 65 is
also decremented by those persons aged 64 and over who will turn 65 (expressed as a factor)
during the estimates cycle. This process of estimating the population aged 64 years old for
counties is achieved through the following steps:

   1. Develop a county-level ratio of the population aged 64 to the population 65 and older in
the last census.

   2. Develop a national-level ratio of the population aged 64 to the population 65 and older in
the last census.

   3. Develop a national-level ratio of the population aged 64 to the population 65 and older for
each estimate year.

   4. Update the county-specific ratios with the change in the national level ratios (estimate
year/census base)

   5. Apply the county-specific ratios to the household population under 65 in the estimates
base.



Components of change for the household population under age 65

TABLE 1--Line 5
Estimated resident births, 7/1 (prior year) to 6/30 (estimate year)--Resident births are recorded by
residence of mother, regardless of where the birth occurred; hence, a county need not
have a hospital to have resident births. If birth data are not available by county for all states for
the estimate year when the county estimates are produced, we assume that prior year county
birth data approximate estimate year births.

TABLE 1--Line 6
Estimated resident deaths to the household population under 65, 7/1 (prior year) to 6/30 (estimate
year)--We use death data tabulated by the most recent residence of the decedent, not by
the place where death occurred. We tabulate NCHS (National Center for Health Statistics)
deaths to the population under 65 years, by race, and control these to state specific tabulations
provided by the FSCPE (Federal State Cooperative Program for Population Estimates) agencies.
We then rake the estimated deaths so they sum to national death totals by race. If estimate
year death data are not available by county for all states when the county estimates are produced,
we assume that prior year county death data approximate estimate year deaths. Our estimate
of deaths for the under 65 years of age population is shown on Line 6.

TABLE 1--Line 7
Estimated net movement from abroad (Immigration), 7/1 (prior year) to 6/30 (estimate year)--We
estimate the number of foreign immigrants who move into the county during the estimate
interval. The estimate is based on the national estimate of foreign migration developed by the
Census Bureau. Our estimate includes emigration from the United States and the immigration of
refugees, legal immigrants, undocumented immigrants, net movement from Puerto Rico, and
federal and civilian citizen movement from abroad. We allocate the national estimate of the
number of undocumented immigrants to states and counties by using the distribution of the
foreign born population who arrived between 1985 and 1990 and were enumerated as residents in
the 1990 census. Legal immigrants and refugees are distributed to counties on the basis of
county of intended residence as reported to the Immigration and Naturalization Service.

TABLE 1--Line 8
Estimated migration base--We develop the migration base by adding one half of the following to
the household base population under 65 years (Line 4): estimated resident births (Line 5), minus
estimated resident deaths under 65 years (Line 6), plus estimated net immigration (Line 7). Only
half of the additions/ deletions to the population would have taken place by the midpoint of the
twelve months. A factor of one-half, or an "exposure factor" must be entered into the equation.
The population base at risk of migrating is usually considered the base at the midpoint of
the period since the population at the beginning of the estimate period has not yet experienced
the births and deaths that are reflected in the population at the end of the period. The
population at the end of the period includes inmigrants and excludes outmigrants. The best
compromise is to take the population at the midpoint of the period. Estimated resident
births, estimated deaths to people under age 65, and net immigration from abroad are assumed to
have been evenly distributed throughout the estimate interval and, therefore, exposed to the
potential for migration, on average, for one-half of the period.

TABLE 1--Line 9
Estimated migration rate--We use an extract of data from Federal income tax returns to measure
the internal, or domestic, migration of the population under 65 years of age, by matching the
returns for
successive years for that age group. We determine the status of the filer by noting the address,
used as a proxy for place of residence, on tax returns filed in the prior year and in the
estimate year. We categorize the filers in each county into: (1) inmigrants (INS), (2) outmigrants
(OUTS), and (3) nonmigrants (NONMOVERS). We derive a net migration rate for each county,
based on the difference between the inmigration and outmigration of tax filers and his or her
dependents.
TABLE 1--Line 10
Estimated net domestic migration--Net migration is the product of the migration base (Line 8)
and the net migration rate (Line 9). If this figure is preceded by a minus sign (-), then the figure
indicates net outmigration; otherwise, the figure represents net inmigration.


Estimated populations

TABLE 1--Line 11
Estimated household population under age 65--The household base population under age 65
(Line 4) is combined with the estimated components of change for the household population
under age 65 to arrive at the estimated household population under age 65 in the estimate year.

TABLE 1--Line 12
Estimated group quarters population under age 65--We assume that military personnel living off
base and those living on base in family quarters are captured in the components of change of the
household population, described above. Military barracks population figures and crews of naval
vessels are obtained from an annual survey of on-base housing facilities for unaccompanied
personnel. The survey is conducted by the Department of Defense (DOD) in September each
year. Certain states also collect military barracks population for their states and those data are
used where appropriate. If Department of Defense survey results are not available in the estimate
year, the prior year results are used as estimates for the current estimate year. College students
living in dormitories, inmates of correctional and juvenile facilities and persons in health care
facilities, nursing homes and Job Corps Centers are also included in this estimate. We exclude
persons aged 65 and over residing in nursing homes and persons in homes for the aged from this
estimate since they are implicitly included in the separate estimate of the 65+ population. Data
on college dormitory populations relate generally to the fall of the preceding year. Data on
institutional populations relate to July 1 of the estimate year or represent some type of average
for the estimate year. Data on the nonmilitary components of the group quarters population are
collected primarily through an annual group quarters report submitted by state members of the
Federal-State Cooperative Program for Population Estimates. The group quarters report is used
to estimate the change in the group quarters population between the prior year and the estimate
year. If no data are available for any component of the group quarters population, we assume
that no change has occurred. In the first estimate year following a decennial census, the group
quarters report is used to estimate the change in group quarters population between the census
and the estimate year.

TABLE 1--Line 13
Estimated total population under age 65--The estimated total population under age 65 is the
estimated household population under age 65 (Line 11) and the estimated group quarters
population under 65 (Line 12).
Explanations of Table 2: Estimate of the Population Aged 65 and Over

Base Populations

TABLE 2--Line 1
Base total population aged 65 and over--The base population for the estimate of the population
age 65 and over is the revised household population estimate of the 65 and over population for
the prior estimate year. We use the county-level tabulations of the number of medicare enrollees
obtained from the Health Care Financing Administration to assist us in developing the estimate.
The availability of this data allows us to develop a separate estimate of change in the population
over age 64. If the Medicare enrollment data are not available at the time we prepare the
estimates, we use the change from the prior estimate year as an estimate of change in the current
estimate.

TABLE 2--Line 2
Base group quarters population aged 65 and over--This component is an estimate of the
population aged 65 and over residing in nursing homes, prisons and other group quarters
facilities.

TABLE 2--Line 3
Estimated population turning 65 in current year--This component is an estimate of the population
who reached their 65th birthday during the estimate year. They are in a sense the number of
people "born" into the 65 and older age group.

TABLE 2--Line 4
Household base population aged 65 and over--This component is calculated by subtracting the
group quarters population (Line 2) and adding the population turning age 65 in the current year
(Line 3).


Components of change for the household population aged 65 and over

TABLE 2--Line 5
Estimated resident deaths to the household population aged 65 and over, 7/1 (prior year) to 6/30
(estimate year)--Our estimate of deaths for the population aged 65 and older is shown on Line 5.
See the explanation of under 65 deaths for more detail.

TABLE 2--Line 6
Estimated foreign migration 65+, 7/1 (prior year) to 6/30 (estimate year)--Same type of
calculation as under 65 foreign immigration (Appendix Table 1, Line 7)

TABLE 2--Line 7
Estimated migration base 65+--Same type of calculation as under 65 migration base (Appendix
Table 1, Line 8)
TABLE 2--Line 8
Estimated migration rate 65+--The migration rate for the population aged 65 and older is
obtained by

   MIGRO={MEDt-[MEDt-1+((AGEt-DEAOt)*MCOV)]}/MEDt-1

where MED is Medicare enrollees, Age is the Population turning 65 in the current year, DEAO is
the Period deaths to the population 65 and over and MCOV is the Medicare Coverage (Medicare
 coverage is defined as Medicare enrollees aged 65 and over in 1990 divided by the Census
population aged 65 and over).

TABLE 2--Line 9
Estimated net migration 65+--Same type of calculation as under 65 net migration (Appendix
Table 1, Line 10)

TABLE 2--Line 10
Estimated household population aged 65 and over--The household base population aged 65 and
over (Line 4) is combined with the estimated components of change for the household population
aged 65 and over to arrive at the estimated household population aged 65 and over in the estimate
year.

TABLE 2--Line 11
Estimated group quarters population aged 65 and over--Persons aged 65 and over residing in
nursing homes, correctional facilities and other group quarters. See under 65 calculation.

TABLE 2--Line 12
Estimated total population aged 65 and over--The estimated total population aged 65 and over is
the sum of the estimated household population aged 65 and over (Line 10) and the estimated
group quarters population 65 and over (Line 11).


Explanations of Table 3: Final Total Population Estimate

TABLE 3--Line 1
Estimated total population under 65--Copied from Appendix Table 1, Line 13

TABLE 3--Line 2
Rake factor for the population under age 65 (Line 2)--The rake factor is used to ensure
consistency between county estimates and independent estimates for the entire population of the
United States. The factor is the national estimate of the total population under age 65 divided by
the sum of the estimated total population under age 65 for all counties in the nation.

TABLE 3--Line 3
Final under age 65 estimate--The final under age 65 population is the estimated total population
multiplied by the rake factor.
TABLE 3--Line 4
Estimated total population aged 65 and over--Copied from Appendix Table 2, Line 12.

TABLE 3--Line 5
Rake factor for the population aged 65 and over--The rake factor is used to ensure consistency
between county estimates and independent estimates for the entire population of the United
States. The factor is the national estimate of the total population aged 65 and over divided by the
sum of the estimated total population aged 65 and over for all counties in the nation.

TABLE 3--Line 6
Final estimate for the population aged 65 and over--The final 65 and older population is the
estimated total population multiplied by the rake factor.

TABLE 3--Line 7
Final total estimate--The final estimate is the sum of the under 65 estimate and the 65 and over
estimate.
APPENDIX B



SUBCOUNTY POPULATION ESTIMATES METHODOLOGY

Explanations of Table 4: Subcounty Population Estimate

Housing Units (Line 1)-- Total housing units from the 1990 Census.

Estimated Residential Construction (Line 2)-- Building permits are compiled from internal data
files developed by Manufacturing and Construction Statistics Division. These files include
imputed permits where a local jurisdiction did not report permit issuance for the entire year.
Housing growth calculated from building permits employs a census region-specific lag time
(weighted average by number of units in structure) between the issuance of permits and
completion of construction. The lag times by census region (in number of months) are:

   U.S. average     7.5
   Northeast Region 8.3          Midwest Region      7.5
   South Region     7.2          West Region         8.7

Two percent of all building permits never result in the actual construction of a housing unit (as
derived from U.S. Bureau of the Census Current Construction Reports, Series C20-9103 and
Series C22-9107). Therefore, a factor of .98 is used to estimate completed new units.

Many jurisdictions provide only annual data to the Building Permit Survey. These reporting
practices have an impact on the ability of the estimation method to measure seasonal fluctuations
in the growth of housing stock. New residential construction was calculated in the following
manner for the July 1, 1998 estimates:

         NC98 = [[(BP89 + BP90 + BP91 + ... BP98] * .98] + NPC98

Where:
         NC98 = Estimate of new residential construction for the period: April 1, 1990 to July 1,
               1998

         BP89 = Residential building permits issued in 1989 that would result in the completion of
               a housing unit after April 1, 1990 and prior to the date of the estimate.

         BP90 - BP98 = Residential building permits issued in 1990-1998 that would result in the
                completion of a housing unit after April 1, 1990 and prior to the date of the
                estimate.

         NPC98 = Estimate of new residential construction occurring in non-permit issuing areas
              for the period: April 1, 1990 to July 1, 1998
Note: Estimates of nonpermitted residential construction (NPC98) are calculated using the
following steps:

At the county level

         1. Calculate the number of 1990 county housing units located in non-permit issuing
         jurisdictions within the county by subtracting the number of units located in permit
         issuing jurisdictions from the total units.

         2. Sum the housing units located in non-permit issuing jurisdictions to the national level.

         3. Calculate the county's share of units located in jurisdictions not issuing permits by
         dividing the county's units by the nation's units -- (1)/(2).

         4. Multiply the total number of units reported in the Survey of Construction as
         constructed without building permits by the county's share of units located in jurisdictions
         not issuing building permits.


At the subcounty level

         5. If a subcounty area was represented by a permit issuing jurisdiction, the number of
         nonpermitted units constructed is set to zero.

         6. If a subcounty area was not represented by a permit issuing jurisdiction, calculate the
         subcounty area's share of the county housing units located in non-permit issuing
         jurisdictions.

         7. Apply the proportion calculated in step 6 to the county estimate of nonpermitted
         construction.


Estimated New Mobile Home Placements (Line 3)-- Tabulations of mobile homes shipped to
states are consistent with Series C-40 reports. Ninety-eight percent of all mobile homes shipped
to states are used for residential purposes. Internal studies have shown that on average a two
month lag time exists between the time when a mobile home is shipped and subsequently put in
place for residential use (Schwanz, 1986). These factors result in the development of the
following formula:.

         NM98 = [(MH90*(11/12)) + MH91 +...+ MH97 + (MH98*(4/12))] * .98

Where:
         NM98 = Estimated new residential mobile home placements since April 1, 1990

         MH90-98 = Mobile homes shipped to states in 1990 - 1998
State mobile home information is distributed to counties based upon a county's proportion of the
state's mobile homes as of the 1990 census.

Estimated Housing Loss (Line 4)-- Demolition permits were compiled from internal data files
developed by Manufacturing and Construction Division (M.D.). These files include imputed
permits where a permit issuing jurisdiction did not report permit issuance for the entire year. No
lag time is assumed for demolition permits.

         HL98 = (.75 * DP90) + DP91 +...DP94 + (.5 * DP95) + DL98

Where:
         HL98 = Estimate of housing loss for the period: April 1, 1990 to July 1, 1993

         DP90-95 = Residential demolition permits issued in 1990-1995

         DL98 = Estimated housing unit loss not measured by demolition permits for the period,
               April 1, 1990 to July 1, 1998

Note: MCD. stopped collecting data on demolition permits in 1995. After 1995, all data on
housing unit loss is estimated. Estimates of nonpermitted demolitions (DL98) are calculated
using the following steps:

         1. Calculate the county's share of structures in region at risk for demolition using data on
         the county's share of the following types of structure:

                1.   Mobile homes and other
                2.   Older units (pre-1939 construction)
                3.   Vacant for Seasonal and Recreational Use
                4.   Boarded up

Data from the Components of Inventory Change (CINCH) study show these units to be at a
greater risk to loss.

         2. Apply the proportion calculated in (1) to CINCH regional loss estimates for
         uninhabitable units.

         3. Apply the proportion calculated in (1) to CINCH regional loss estimates for
         demolitions and disasters.

         4. If specific housing unit loss data due to disaster are known, override step 3.

         5. For each year since 1990, rake the difference between the estimated regional total of
         units lost to disaster and demolitions and the CINCH regional loss data to disaster and
         demolitions, to all subcounty areas within the region.

         6. For each subcounty area, subtract the number of housing units lost to demolitions and
         disasters as reported on demolition permits from the estimate of housing unit loss to
         demolition and disaster not covered by demolition permits. If the difference is zero or
         less, all loss due to demolitions and disaster is covered by demolition permits.
         7. Total housing unit loss, not covered by demolition permits, for any subcounty area is
         the sum of the estimate of units becoming uninhabitable from (2) and the estimate of loss
         due to demolition and disaster not covered by demolition permits from (6).

Note: In essence, this process operates by applying a crude death rate to the housing inventory to
develop a baseline estimate of housing unit loss. This crude death rate is refined by the type
of housing stock contained within each subcounty area. After application of the crude death rate,
if real data exists (demolition permits) for the subcounty area, that data is used as a
sub-component of loss. The sum of permitted and non-permitted housing loss exceeds the
baseline estimate of housing unit loss only in cases where permitted demolitions, or known
disaster losses are greater than the estimate developed by the crude death rate.

Housing Units in Estimate Year (Line 5)-- Housing unit estimates for each area were developed
by the component model described below. The July 1, 1996 estimates are used here as an
example.

         HU98 = HU90 + (NC98 + NM98) - HL98

Where:

         HU98 = Estimated 1996 housing units

         HU90 = 1990 census housing units

         NC98 = Estimated residential construction, April 1, 1990 to July 1, 1998

         NM98 = Estimated new residential mobile home placements, April 1, 1990 to
         July 1, 1998

         HL98 = Estimated residential housing loss, April 1, 1990 to July 1, 1998

1990 Occupancy Rate (Line 6)-- The occupancy rate is calculated by dividing the number of
occupied units in the locality by the total number of units as reported in the 1990 Census.

1990 Persons per Household (Line 7)-- The number of persons per household is obtained by
dividing the household population as reported in the 1990 Census by the number of occupied
housing units in 1990.

Uncontrolled Household Population (Line 8)-- The uncontrolled household population is
obtained by multiplying the number of housing units in the estimate year by the occupancy rate
and the persons per household.
Uncontrolled County Household Population Estimate (Line 9)-- The uncontrolled county
population is the sum of the uncontrolled subcounty area household population estimates within
the county.

Published County Estimate (Line 10)-- The published county population estimate as calculated
by the tax return method for the current estimate year.

County Rake Factor (Line 11)-- The published county population estimate divided by the
uncontrolled county population estimate.

Controlled Household Population Estimate (Line 12)-- The uncontrolled household population
estimate multiplied by the county rake factor.

Group Quarters Population (Line 13)-- This component is primarily a combination of military
personnel living in barracks, college students living in dormitories and persons residing in
institutions. Inmates of correctional and juvenile facilities and persons in health care facilities
and Job Corps Centers are also included in this category.

Final Population Estimate (Line 14)-- The sum of the Controlled household population estimate
and the group quarters population estimate.




.

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:18
posted:4/3/2011
language:English
pages:24