Methodology for Constructing Estimates of Total Population for Counties and Subcounty Areas in Florida

Richard Doty, Suzanne Roulston-Doty, and Stefan Rayer

Bureau of Economic and Business Research

University of Florida

October, 2021

The Bureau of Economic and Business Research (BEBR) makes population estimates for every county and subcounty area in Florida, with subcounty areas defined as incorporated cities, towns and villages, and the unincorporated balance of each county. County estimates are calculated as the sum of the subcounty estimates for each county, and the state estimate is calculated as the sum of the county estimates. The estimates refer solely to permanent residents of Florida; they do not include seasonal or other types of temporary residents.

The estimates are produced using the housing unit method, in which changes in population are based on changes in occupied housing units (or households). This is the most commonly used method for making local population estimates in the United States because it can utilize a wide variety of data sources, can be applied at any level of geography, and can produce estimates that are at least as accurate as those produced by any other method.

The foundation of the housing unit method is the fact that almost everyone lives in some type of housing structure, whether a traditional single family unit, an apartment, a mobile home, a college dormitory, or a state prison. The population of any geographic area can be calculated as the number of occupied housing units (households) times the average number of persons per household (PPH), plus the number of persons living in group quarters such as college dormitories, military barracks, nursing homes, and prisons:

Pt = (Ht x PPHt) + GQt

where Pt is the population at time t, Ht is the number of occupied housing units at time t, PPHt is the average number of persons per household at time t, and GQt is the group quarters population at time t. Estimates of the number of people without permanent living quarters (e.g., the homeless population) are included in estimates of the group quarters population.

This is an identity, not an estimate. If these three components were known exactly, the total population would also be known. The problem, of course, is that these components are almost never known exactly. Rather, they must be estimated from various data sources, using one or more of several possible techniques. In this report, we describe the data and techniques used to develop population estimates for Florida’s counties and subcounty areas for April 1, 2021.

Households
Census definitions require a person to be counted as an inhabitant of his/her usual place of residence, which is generally construed to mean the place where she/he lives and sleeps most of the time. This place is not necessarily the same as one’s legal or voting residence. A household is the person or group of people occupying a housing unit; by definition, the number of occupied housing units is the same as the number of households. Households refer solely to permanent residents and a housing unit is classified as vacant even when it is continuously occupied, if all the occupants are temporary residents staying only for a few days, weeks, or months.

BEBR uses three different data sources to estimate the number of households in Florida. Our primary data source is active residential electric customers. We collect these data from each of the state’s 54 electric utility companies. Households can be estimated by constructing a ratio of households to active residential electric customers using data from the most recent census year (e.g., 2020) and multiplying that ratio times the number of active residential customers in some later year (e.g., 2021). This procedure assumes that no changes have occurred in electric company bookkeeping practices, in the vacancy rate of active residential electric customers, or in the proportion of those customers who are permanent residents. Although changes do occur, they are generally fairly small. In some places we adjust the household/electric customer ratio to account for changes in the vacancy rate or the proportion of housing units occupied by permanent residents.

We sometimes filter electric customer data to exclude minimum use customers. Minimum use customers are those using less than 200 kilowatt (kWh) hours per month. We believe these customers represent seasonal or other part-time residents or vacant units, and excluding them may give a more accurate measure of permanent residents. Because we estimate the change in population since the 2020 Census, excluding minimum use customers can capture changes in unit occupancy over that period. These data are not available for all areas of the state, but in places in which the data are available and appear to be reliable we may use them in conjunction with other data sources.

Our second data source is residential building permits, as collected and distributed by the U.S. Department of Commerce. The housing inventory in 2021 for a city or county that issues building permits can be estimated by adding permits issued since 2020 to the units counted in the 2020 census and subtracting units lost to destruction, demolition, or conversion to other uses. The time lag between the issuance of a permit and the completion of a unit is assumed to be three months for single-family units and fifteen months for multifamily units. Building permits are not issued for mobile homes, but proxies can be derived from records of shipments to mobile home dealers in Florida. Creating a housing inventory for an entire county requires complete permit data for every permitting agency within the county. Although such data are not always available, coverage is sufficient in most Florida cities and counties to provide useful information.

There are no readily available data sources providing comprehensive up-to-date information on occupancy rates that are as reliable as those produced by the latest decennial census. Accurate information can be obtained through special censuses or large sample surveys, but in most instances these methods are too expensive to be feasible. A common solution is to use the occupancy rates reported in the most recent decennial census. We base our occupancy estimates on these values, but we may make adjustments to account for factors reflecting changes in occupancy rates over time. Occupancy changes since 2020 may be captured in places where we use electrical customer data and are able to exclude minimum use customers. Additional factors may include data from the U.S. Census Bureau’s American Community Survey (ACS), in cases where it shows statistically significant trends over time since the last decennial census.

The product of the inventory figure and the occupancy rate provides an estimate of the number of households. There are several potential problems with this estimate. Time lags between the issuance of permits and the completion of units may vary from place to place and from year to year. The proportion of permits resulting in completed units is usually unknown. Data on demolitions and conversions are incomplete and data on mobile homes must be estimated indirectly. Reliable estimates of changes in occupancy rates are generally unavailable. Certificate-of-occupancy data can eliminate problems related to completion rates and time lags but not those related to occupancy rates, demolitions, and conversions. Although these problems limit the usefulness of the data in some places, building permit data often provide reasonably accurate estimates of households.

Our third data source for estimates is the number of homestead exemptions by county reported by the Florida Department of Revenue. Households can be estimated by constructing a ratio of households to exemptions using data from the most recent census year (e.g., 2020) and multiplying that ratio times the number of exemptions in some later year (e.g., 2021). An important advantage of these data is that they cover only housing units occupied by permanent residents, thereby excluding the impact of seasonal and other non-permanent residents. The primary disadvantage is that the data do not include households occupied by renters or other non-homeowners, but those households often change at a similar rate to the households with homestead exemptions. Homestead exemption data is also available from each county’s property appraiser at the property parcel level, which can be summarized by subcounty areas. We occasionally use these data to inform our decision making in places where our other primary data sources show significantly different results.

Electric customer, building permit, and homestead exemption data all provide useful information regarding changes in households. Previous research on BEBR population estimates has shown that household estimates based on electric customer data are—on average—more accurate than those based on building permit and other data. However, we use our professional judgment to decide which data source(s) to use in each specific county and subcounty area. In many instances, we use averages of estimates from more than one data source. We also sometimes use GIS-based property parcel data (along with year built information and detailed land use codes from the Florida Department of Revenue) to evaluate which data source is best for a particular place.

Persons per Household
The second component of the housing unit method is the average number of persons per household (PPH). Florida’s PPH dropped steadily from 3.22 in 1950 to 2.46 in 1990 but then leveled off, remaining constant between 1990 and 2000 before rising to 2.48 in 2010. There is a substantial amount of variation among local areas in Florida, with values in 2010 ranging from 2.1 to 3.1 for counties and from less than 1.4 to more than 4.0 for subcounty areas. PPH values have risen over time in some cities and counties and declined in others.

For each county and subcounty area, we base our PPH estimates on the local PPH value in the most recent census (e.g., 2010). In some instances, we estimate changes in PPH since that census using statistically significant trends in data from the American Community Survey or changes in the mix of single-family, multifamily, and mobile home units since the last census. Again, we use our professional judgment to decide which data sources and techniques to use in each county and subcounty area.

Group Quarters Population
The household population is calculated as the product of households and PPH. To obtain an estimate of the total population, we must add an estimate of the group quarters population. In most places, we estimate the group quarters population by assuming that it accounts for the same proportion of total population in 2020 as it did in 2010. For example, if the group quarters population accounted for 2% of the total population in 2010, we assume that it accounted for 2% in 2020. In places where there are large group quarters facilities, we collect data directly from the administrators of those facilities and add those estimates to the other group quarters population. Inmates in state and federal institutions are accounted for separately in all local areas; these data are available from the Federal Bureau of Prisons, the Florida Department of Corrections, the Florida Department of Veteran Affairs, the Florida Agency for Persons with Disabilities, the Florida Department of Health, the Florida Department of Juvenile Justice and the Florida Department of Children and Families. The total population estimate is made by adding the estimate of the group quarters population to the estimate of the household population.

Conclusion
The population estimates produced by BEBR are calculated by multiplying the number of households by the average number of persons per household and adding the number of persons living in group quarters. This methodology is conceptually simple but effective. It utilizes data that are available for all local areas, its components respond rapidly to population movements, and it can be applied systematically and uniformly everywhere in the state. A comparison of population estimates with census results for 1980, 1990, 2000, and 2010 showed the BEBR estimates to be quite accurate, especially when compared to other sets of estimates. We believe the housing unit method is the most effective method for making city and county population estimates in Florida and that it produces reliable estimates that provide a solid foundation for budgeting, planning, and analysis.

Acknowledgement
Funding for these estimates was provided by the Florida Legislature.

Methodology for Constructing Projections of Florida Population by County, 2025–2050, with Estimates for 2021

Stefan Rayer and Ying Wang

Bureau of Economic and Business Research

University of Florida

February 2022

The Bureau of Economic and Business Research (BEBR) has been making population projections for Florida and its counties since the 1970s. This report presents our most recent set of projections and describes the methodology used to construct those projections. To account for uncertainty regarding future population growth, we publish three series of projections. We believe the medium series is the most likely to provide accurate forecasts in most circumstances, but the low and high series provide an indication of the uncertainty surrounding the medium series. It should be noted that these projections refer solely to permanent residents of Florida; they do not include tourists or seasonal residents.

State Projections
The starting point for the state-level projections was the decennial census count for April 1, 2020. Because the detailed census counts by age and sex are not yet available, we used the BEBR age and sex estimates for April 1, 2020, which were controlled to the Census 2020 count of total population. Projections were made in one-year intervals using a cohort-component methodology in which births, deaths, and migration are projected separately for each age-sex cohort in Florida. We applied three different sets of assumptions to provide low, medium, and high series of projections. Although the low and high series do not provide absolute bounds on future population change, they provide a reasonable range in which Florida’s future population is likely to fall.

Survival rates were applied by single year of age and sex to project future deaths in the population. These rates were based on Florida Life Tables for 2012–2018, using mortality data published by the Office of Vital Statistics in the Florida Department of Health. We adjusted the survival rates for 2020–2026 to make them consistent with recent mortality trends, and to align the projected deaths with those from the State of Florida’s Demographic Estimating Conference (DEC) held December 13, 2021. After 2026, we made small adjustments to the survival rates based on projected changes in survival rates released by the U.S. Census Bureau. We used the same mortality assumptions for all three series of projections.

Domestic migration rates by age and sex were based on Public Use Microdata Sample (PUMS) files from the 2011–2019 American Community Survey (ACS) 1-year estimates and 2015–2019 ACS 5-year estimates. We calculated an average of those two sets of migration estimates; projections based on input data from more than one time period tend to be more accurate than those based on a single time period. By combining 1-year ACS estimates, which are more current, with 5-year ACS estimates, which are more stable, we make use of the different strengths of each type of ACS data.

We applied smoothing techniques to the age/sex-specific migration rates to adjust for data irregularities caused by small sample size. The smoothed in- and out-migration rates were weighted to account for recent changes in Florida’s population growth rates. Projections of domestic in-migration were made by applying weighted in-migration rates to the projected population of the United States (minus Florida), using the most recent set of national projections produced by the U.S. Census Bureau. Projections of out-migration were made by applying weighted out-migration rates to the Florida population. In both instances, rates were calculated separately for males and females for each age up to 90 and over.

For the medium projection series, in-migration weights for total population varied from 1.26 to 1.01, and out-migration weights varied from 0.97 to 1.00. For the low projection series, the in-migration weights described above were lowered over time – from 7.6% in 2022 to 11% in 2050; the out-migration weights were raised by the same margins. For the high projection series, the in-migration weights described above were raised over time – from 7.6% in 2022 to 11% in 2050; the out-migration weights were lowered by the same margins.

The distribution of foreign immigrants by age and sex was also based on averages of the patterns observed over the same time periods using the same ACS data sets as for domestic migration. Again, we smoothed the estimates to account for irregularities in the age/sex distribution of immigrants. For the medium projection series, we held foreign immigration at an average of the observed levels, with some short-term adjustments based on recent trends. For the low series, foreign immigration was projected to decrease by 2,900 per year from the average of the observed levels; for the high series, foreign immigration was projected to increase by 2,500 per year. Foreign emigration was assumed to equal 25% of foreign immigration for each series of projections.

Projections were made in one-year intervals, with each projection serving as the base for the following projection. Projected in-migration for each one-year interval was added to the survived Florida population at the end of the interval and projected out-migration was subtracted, giving a projection of the population age one and older.

Births were projected by applying age-specific birth rates (adjusted for child mortality) to the projected female population. These birth rates were based on Florida birth data for 2012–2018 published by the Office of Vital Statistics in the Florida Department of Health. They imply a total fertility rate (TFR) of 1.75 births per woman for total population. These rates were reduced in the short-term projections to about 1.66 births per woman to make them consistent with recent fertility trends, and to align the projected births with those from the December 13, 2021 DEC. After 2026, we raised birth rates gradually; the projections from 2034 to 2050 imply about 1.78 births per woman.

The medium projections of total population for 2022–2026 were adjusted to be consistent with the state population forecasts for those years produced by the December 13, 2021 DEC. None of the projections after 2026 had any further controls.

County Projections
The cohort-component method is a good way to make population projections at the state level but is not necessarily the best way to make projections at the county level. Many counties in Florida are so small that the number of persons in each age-sex category is inadequate for making reliable cohort-component projections, given the lack of detailed small-area data. Even more important, county growth patterns are so volatile that a single technique based on data from a single time period may provide misleading results. We believe more useful projections of total population can be made by using several different techniques and historical base periods.

For counties, we started with the population estimate constructed by BEBR for April 1, 2021. We made projections for each county using five different techniques in five-year increments. The five techniques were:

1. Linear – the population will change by the same number of persons in each future year as the average annual change during the base period.
2. Exponential – the population will change at the same percentage rate in each future year as the average annual rate during the base period.
3. Share-of-growth – each county’s share of state population growth in the future will be the same as its share during the base period.
4. Shift-share – each county’s share of the state population will change by the same annual amount in the future as the average annual change during the base period.
5. Constant-share – each county’s share of the state population will remain constant at its 2021 level.

For the linear and share-of-growth techniques we used base periods of two, ten, and twenty years (2019–2021, 2011–2021, and 2001–2021), yielding three sets of projections for each technique. For the exponential and shift-share techniques we used base periods of five and fifteen years (2016–2021 and 2006–2021), yielding two sets of projections for each technique. The constant-share method was based on data for a single year (2021).

This methodology produced eleven projections for each county for each projection year (2025, 2030, 2035, 2040, 2045, and 2050). From these, we calculated five averages: one using all eleven projections (AVE-11), one that excluded the highest and lowest projections (AVE-9), one that excluded the two highest and two lowest projections (AVE-7), one that excluded the three highest and three lowest projections (AVE-5), and one that excluded the four highest and four lowest projections (AVE-3). Based on the results of previous research, we designated the average that excluded the three highest and three lowest projections (AVE-5) as the default technique for each county. We evaluated the resulting projections by comparing them with historical population trends and with the level of population growth projected for the state as a whole. For counties in which AVE-5 did not provide reasonable projections, we selected the technique producing projections that fit most closely with our evaluation criteria.

For 56 counties we selected AVE-5, the average in which the three highest and three lowest projections were excluded. In the remaining 11 counties, we selected projections made from an individual technique or calculated a custom average (e.g., an average of two individual techniques). These include Bay, Calhoun, Gadsden, Glades, Hardee, Holmes, Jackson, Liberty, Madison, Monroe, and Okeechobee counties.

We also made adjustments in several counties to account for changes in institutional populations such as university students and prison inmates. Adjustments were made only in counties in which institutional populations account for a large proportion of total population or where changes in the institutional population have been substantially different than changes in the rest of the population. In the present set of projections, adjustments were made for Alachua, Baker, Bradford, Calhoun, Columbia, DeSoto, Dixie, Franklin, Gadsden, Gilchrist, Glades, Gulf, Hamilton, Hardee, Hendry, Holmes, Jackson, Jefferson, Lafayette, Leon, Liberty, Madison, Okeechobee, Santa Rosa, Sumter, Suwannee, Taylor, Union, Wakulla, Walton, and Washington counties.

Range of County Projections
The techniques described in the previous section were used to construct the medium series of county projections. This is the series we believe will generally provide the most accurate forecasts of future population change. We also constructed low and high projections to provide an indication of the uncertainty surrounding the medium county projections. The low and high projections were based on analyses of past population forecast errors for counties in Florida, broken down by population size and growth rate. They indicate the range into which approximately three-quarters of future county populations will fall, if the future distribution of forecast errors is similar to the past distribution.

The range between the low and high projections varies according to a county’s population size in 2021 (less than 30,000; 30,000 to 199,999; and 200,000 or more), rate of population growth between 2011 and 2021 (less than 7.5%; 7.5–15%; 15–30%; and 30% or more), and the length of the projection horizon (on average, projection errors grow with the length of the projection horizon). Our studies have found that the distribution of absolute percent errors tends to remain fairly stable over time, leading us to believe that the low and high projections provide a reasonable range of errors for most counties. It must be emphasized, however, that the actual future population of any given county could be below the low projection or above the high projection.

For the medium series of projections, the sum of the county projections equals the state projection for each year (except for slight differences due to rounding). For the low and high series, however, the sum of the county projections does not equal the state projection. The sum of the low projections for counties is lower than the state’s low projection and the sum of the high projections for counties is higher than the state’s high projection. This occurs because potential variation around the medium projection is greater for counties than for the state as a whole.

Note
For this set of population projections, we did not make specific adjustments related to the ongoing COVID-19 pandemic. The estimated statewide population growth from April 1, 2020 to April 1, 2021 of about 360,000 persons was comparable to annual population changes in the late 2010s. Furthermore, the most recent state projections from the December 13, 2021 DEC, to which these county projections are controlled, show similar statewide growth over the next five years as the state projections adopted at the December 3, 2019 DEC before the pandemic. Consequently, while the pandemic has to some extent impacted the components of Florida’s population change – especially natural increase, which has been negative since 2020 – we currently expect no particular changes to the projected population levels for 2025 and beyond.

Acknowledgement
Funding for these projections was provided by the Florida Legislature.

Methodology for Constructing Population Projections by Age, Sex, Race, and Hispanic Origin for Florida and Its Counties, 2025–2045, With Estimates for 2020

Stefan Rayer and Ying Wang

Bureau of Economic and Business Research

University of Florida

June, 2021

The composition of Florida’s population has changed substantially in recent decades. Between 1950 and 2010, for example, the proportion of Florida’s population younger than age 15 declined from 26.2 to 17.5 percent; the proportion age 65 and older rose from 8.6 to 17.3 percent; and the proportion black declined from 21.7 to 16.9 percent. The Hispanic population increased from 6.0 percent of the total population in 1970 to 22.5 percent in 2010. Changes in demographic composition have been even greater for many counties than for the state as a whole.

These changes have important implications for planning and public policy. They affect the demand for education, healthcare, housing, recreation, transportation, and many other goods and services. They affect the number and characteristics of persons in the labor force and in public and private retirement systems. They affect the allocation of many types of public funds. Consequently, there is a tremendous need for population estimates and projections by age, sex, race, and Hispanic origin.

We note that this set of projections is still based on the Census 2010 counts and the BEBR population estimates since then. The next set of BEBR county projections by age, sex, race, and Hispanic origin – scheduled for release in 2022 – will incorporate the Census 2020 counts.

Definitions of Race and Ethnicity
The decennial census in the United States is based on self-enumeration. Residents of each household are asked to provide the responses they believe best describe their demographic characteristics, based on guidelines established by the U.S. Census Bureau and the U.S. Office of Management and Budget (OMB). These guidelines allow respondents to identify themselves as Hispanic or non-Hispanic and as belonging to one or more of several racial groups.

It should be noted that “Hispanic” is an ethnic classification rather than a racial category; that is, people can be identified both by Hispanic origin and by race. The OMB defines Hispanic or Latino as “a person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin regardless of race.” For data collection and presentation purposes, federal agencies are required to use a minimum of two ethnicities: “Hispanic or Latino,” and “Not Hispanic or Latino.” We follow the same guidelines in this report and use the term “Hispanic” to refer to persons of Hispanic, Latino, or Spanish origin.

The three largest racial/ethnic groups in Florida are non-Hispanic whites, non-Hispanic blacks, and Hispanics. These three groups accounted for 97.0 percent of Florida’s population in 2010. We made an initial set of estimates and projections by age and sex for these three racial/ethnic groups. Using these estimates and projections as a starting point, we constructed additional sets of estimates and projections for several other race/ethnicity combinations.

DATA
Data from the 1990, 2000, and 2010 censuses formed the basis for these estimates and projections. Although census data are generally quite reliable, two issues regarding race and ethnicity complicate their use. First, in 1990 and all previous censuses, respondents were required to identify themselves as belonging to a single race. Starting in 2000, they were permitted to identify themselves as belonging to one or more races. In Florida, 97.5 percent of the population identified themselves as belonging to a single race in 2010 and 2.5 percent identified themselves as belonging to two or more races. These proportions are very similar to those reported in 2000 (97.6 percent and 2.4 percent, respectively).

Second, although the U.S. Census Bureau defines “Hispanic origin” as an ethnic classification rather than a racial category, many respondents interpreted it as a racial category and listed their race as Hispanic, Latino, Mexican, Spaniard, or a similar response. In Florida, 3.6 percent of the total population in 2010 classified themselves as belonging to some race other than those listed on the census questionnaire; more than 90 percent of those respondents were of Hispanic origin. In 2000, 4.4 percent of the total population classified themselves as belonging to some race other than those listed on the census questionnaire; again, more than 90 percent of those respondents were of Hispanic origin.

Responding to these issues, the National Center for Health Statistics (NCHS) collaborated with the U.S. Census Bureau to create a set of modified 2000 census counts for every state and county in the United States. Using a variety of data sources and techniques, the NCHS assigned people who classified themselves as belonging to more than one race (or who marked “some other race” on the census questionnaire) to a single primary race. These modifications produced data that were consistent over time, prevented double-counting of people belonging to more than one race, and provided a racial classification for Hispanics who did not identify their race. The NCHS released a comparable set of modified census counts for 2010. For 1990, the U.S. Census Bureau made modifications to the 1990 census counts in which persons listing “some other race” were assigned to a specific race – the modified age/race, sex, and Hispanic origin (MARS) files. The estimates and projections described in this report were based on MARS data for 1990 and NCHS data for 2000 and 2010. For 2010, we used an updated April 1, 2010 population that includes Count Question Resolution (CQR) changes for Lake, Marion, and Miami-Dade counties from the Vintage 2013 NCHS bridged-race postcensal population estimates. We also made an additional adjustment for Gadsden County to correct for the institutional facility that the 2010 census failed to enumerate in the county.

Large institutions (e.g., universities, prisons) account for a significant proportion of the total population in many counties in Florida. In such counties, it is important to account for the impact of these institutions when making population estimates and projections. Consequently, we used institutional records and data from the decennial census to estimate the non-institutional population by age, sex, race, and Hispanic origin for 1990, 2000, 2010, and 2020 in the following counties: Alachua, Baker, Bradford, Calhoun, Columbia, DeSoto, Dixie, Franklin, Gadsden, Gilchrist, Glades, Gulf, Hamilton, Hardee, Hendry, Holmes, Indian River, Jackson, Jefferson, Lafayette, Leon, Liberty, Madison, Okeechobee, Santa Rosa, Sumter, Suwannee, Taylor, Union, Volusia, Wakulla, Walton, and Washington. In these counties, we made separate projections for the institutional and non-institutional populations. The final estimates and projections for each county were constructed by adding together the institutional and non-institutional populations. The remainder of this report describes the methodology used for making estimates and projections of the non-institutional population.

METHODOLOGY
2020 Estimates of Total Population by Race and Ethnicity
We made estimates of the total number of non-Hispanic whites, non-Hispanic nonwhites, and Hispanics for 2020 using a variety of data sources and techniques. Some relied on extrapolations of previous population trends, whereas others incorporated data on births, deaths, and school enrollment by race and ethnicity. Some estimates were based on averages of several of the individual techniques. The final estimate for each racial/ethnic group in each county was based on our judgment regarding which technique was most likely to provide an accurate estimate of the non-institutional population. Estimates of total population by race and ethnicity were made by adding estimates of the institutional population to estimates of the non-institutional population. As a final step, estimates for the three racial/ethnic groups were controlled to the 2020 estimates of total population published in “Florida Estimates of Population: April 1, 2020,” Bureau of Economic and Business Research, December 2020. A more detailed description of the methodology can be found in an article by Stanley Smith and June Nogle published in the Social Science Quarterly in 2004 (volume 85, pp. 731–745).

Projections of Total Population by Race and Ethnicity
Starting with the 2020 estimates, we made projections of the total non-institutional population of each county for non-Hispanic whites, non-Hispanic nonwhites, and Hispanics using the following techniques:

LINE20: linear extrapolation of 2000–2020 non-institutional population change for each racial/ethnic group.

LINE10: linear extrapolation of 2010–2020 non-institutional population change for each racial/ethnic group.

SHARE20: each racial/ethnic group’s share of county non-institutional population change 2000–2020 is applied to projected county non-institutional population change.

SHARE10: each racial/ethnic group’s share of county non-institutional population change 2010–2020 is applied to projected county non-institutional population change.

EXPO AVE: average of three exponential extrapolations of 1990–2020, 2000–2020, and 2010–2020 non-institutional population change for each racial/ethnic group.

SHIFT AVE: average of the three changes in each racial/ethnic group’s share of county non-institutional population 1990–2020, 2000–2020, and 2010–2010, which are linearly extrapolated and applied to county projections of total non-institutional population.

CONST%: each racial/ethnic group’s share of the non-institutional population in 2020 is assumed to remain constant over time.

AVE7: an average of projections from the seven techniques described above.

AVE5: an average of these projections, excluding the highest and lowest.

AVE3: an average of these projections, excluding the two highest and the two lowest.

CTRL AVE7: AVE7 controlled to medium county projection of total non-institutional population.

CTRL AVE5: AVE5 controlled to medium county projection of total non-institutional population.

CTRL AVE3: AVE3 controlled to medium county projection of total non-institutional population.

The final projection of the total population for each racial/ethnic group in each county was based on our judgment regarding which technique was most likely to provide an accurate forecast of the future non-institutional population. In 62 counties, the final projection was based on CTRL AVE3. In the remaining five counties, we selected projections made from an individual technique or calculated a custom average (e.g., an average of two individual techniques, or an average of one individual technique and one of the above described averages). These include Gadsden, Glades, Jackson, Monroe, and Putnam counties.

In counties with institutional adjustments, projections of the institutional population were based on institutional records and our judgment regarding future institutional growth. Projections of the racial/ethnic breakdown of the institutional population were made by applying the racial/ethnic distribution from the 2010 census to the projections of the total institutional population, which were adjusted to reflect changes in the racial/ethnic distribution of the non-institutional population over the projection horizon.

Finally, projections of total population by race/ethnicity were made by adding projections of the institutional population to projections of the non-institutional population. In all counties, projections for the three racial/ethnic groups were controlled to the medium projections published in “Projections of Florida Population by County, 2025–2045, with Estimates for 2020,” Florida Population Studies, Bulletin No. 189, Bureau of Economic and Business Research, April 2021.

Estimates and Projections by Age, Sex, and Race/Ethnicity
Estimates and projections by age and sex for each of the three racial/ethnic groups were made using a cohort-survival rate methodology. Age was calculated in five-year groups from 0–4 to 85+. Estimates and projections were made in five-year intervals, starting with the 2015 estimates published in June 2016; the 2020 estimates served as the base for the following projections.

Using modified census and institutional population data for 2000 and 2010, and intercensal population estimates and institutional population data for 2005, we subtracted the institutional population from the total population for each age, sex, racial, and ethnic group to derive estimates of the non-institutional population in each demographic subgroup. We calculated cohort-survival rates by sex for the non-institutional population by dividing the 2010 modified census count for each age, racial, and ethnic group by the 2005 intercensal population estimate for the corresponding group 5 years younger. We also calculated cohort-survival rates by sex for the non-institutional population by dividing the 2005 intercensal population estimate for each age, racial, and ethnic group by the 2000 modified census count for the corresponding group 5 years younger. From these we calculated an average of 2000–2005 and 2005–2010 cohort-survival rates. We chose an average of those two periods because population growth in the first half of the decade was quite different from population growth in the second half. Averaging has generally been found to increase the accuracy of population estimates and projections.

Using cohort-survival rates averaged over 2000–2005 and 2005–2010, we made several additional adjustments. First, we applied weighting factors to account for higher survival rates among the older age groups. For many counties, we further adjusted the resulting cohort-survival rates to account for apparent data errors and to smooth out differences among age groups, or between males and females. These adjustments were most frequent in counties with small populations, especially for the non-Hispanic nonwhite and the Hispanic populations.

We applied the adjusted cohort-survival rates to the 2015 non-institutional population by age, sex, race, and ethnicity to produce estimates for 2020 for the population age 5 and older. For the population less than age 5, we used child-woman ratios based on 2010 NCHS data (i.e., population aged 0–4 divided by females aged 15–44). We applied those ratios to the estimated female population in 2020 to provide estimates of children aged 0–4. The population age 0–4 was divided between males and females using proportions of 0.51 and 0.49, respectively. In some instances, we adjusted the child-woman ratios to account for expected changes in fertility rates. For each of the three racial and ethnic groups, we controlled the non-institutional age and sex estimates to the independent estimates of the total non-institutional population for 2020.

We repeated the process to produce projections for 2025, 2030, 2035, 2040, and 2045. These projections were controlled to the independent projections of the non-institutional population described above. As a final step, we added the independent projections of the institutional population, providing projections by age and sex for non-Hispanic whites, non-Hispanic nonwhites, and Hispanics. Projections at the state level were calculated by adding up the county projections.

Estimates and Projections for other Racial/Ethnic Groups
We developed estimates and projections for several additional racial/ethnic groups. Using the 2010 NCHS data, we calculated the white/nonwhite proportion of the Hispanic population for each county and applied those proportions to the Hispanic estimates and projections to provide a white/nonwhite breakdown of the Hispanic population (in Florida, approximately 76 percent of the Hispanic population identified themselves as white alone in the 2010 census). Adding the Hispanic white population to the non-Hispanic white population provided estimates and projections of the total white population by age and sex for each county.

Using the 2010 NCHS data, we calculated blacks as a proportion of nonwhites for both the Hispanic and non-Hispanic populations. We made those calculations separately for each county and – based on historical trends and the 2010 values – projected those proportions into the future. By applying these proportions to estimates and projections of the nonwhite population (for both Hispanics and non-Hispanics), we developed estimates and projections of the non-Hispanic black population and the total black population by age and sex for each county.

ACKNOWLEDGEMENT
Funding for these estimates and projections was provided by the Florida Legislature.