Daniel Sheya prepared this report and the city estimates. Linda Gage prepared the state and county estimates, Walter Schwarm and Doug Kuczynski provided analytical expertise in city estimates while producing military, group quarters, household population and housing estimates. Mary Heim, Chief of the Demographic Research Unit, and John Malson, Research Manager over the city estimates section provided general direction.
State of California, Department of Finance, E-4 Historical Population Estimates for City, County and the State, 1991-2000, with 1990 and 2000 Census Counts. Sacramento, California, August 2007.
- E-4 Historical Population Estimates for City, County and the State, 1991-2000, with 1990 and 2000 Census Counts
This report provides revised population estimates for the state, counties and cities for January 1, 1991 through January 1, 2000. It also includes 1990 and 2000 decennial census counts. The revised estimates attempt to provide a consistent data series reflecting both decennial census counts by utilizing the Error of Closure (EOC) adjustment procedure, which is discussed below.
During the 1990s, California grew by roughly 1.19 percent per year, adding 374,079 persons annually on average. In comparison, California averaged an annual growth rate of 2.6 percent and increased by 609,200 per annum from 1980 to 1990.
Overview. The state, county and city January series estimates are independently estimated using population change models benchmarked on 1990 and 2000 decennial census counts that have not been adjusted for estimated undercount. The state population is estimated using the Driver License Address Change method. County population proportions are estimated using the average of three separately estimated sets of proportions, as described below, and then the final distribution of county proportions is applied to the independently estimated state control to derive the county estimates. These estimates, in turn, are used as controls for the city estimates, which have first been produced using the Housing Unit method detailed below.
Error of Closure. Revising the estimates required computing the differences between “test” city, county and state estimates for April 1, 2000 and actual April 1, 2000 city, county and state census counts, and then allocating these differences, called the “error” statistically to the original estimates. The revisions correct for estimate error in the original series. To view a more elaborate explanation of the Error of Closure technique, please see the “E-8 Historical Population and Housing Estimates” report spanning the period 1990-2000.
City and Unincorporated Area Estimates. The Demographic Research Unit (DRU) uses the Housing Unit (HU) Method to estimate total and occupied housing units, household size and population, and group quarter population. Annual HU change is estimated by adding new construction and annexed housing units and subtracting demolitions from the previous year’s housing totals. The Census Bureau and various local jurisdictions supply these HU changes. The DRU-derived civilian vacancy rate based on 1990 census data is used to estimate civilian households, to which the occupied military units are added to calculate occupied HUs. Military changes, including base realignments and closures, are tracked using surveys. Multiplying the current persons per household estimate by the number of occupied HUs yields an estimate of household population. DRU updates the census group quarter population using reported population in group quarter facilities, and calculating group quarters change. The sum of estimates of the household and group quarter populations equals estimates of total population.
County Estimates. County population proportions result from averaging three methods1:
DLAC Method. A modified version of the state Driver License Address Change (DLAC) method is used for counties. County proportions of the state total result from changes in county population values for births, deaths, school enrollment, foreign and domestic migration, medical aid enrollments, and group quarters population.
Ratio-Correlation Method. This method models change in household population as a function of changes in the distributions of driver licenses, school enrollments, housing units, and deaths. Estimates of county group quarters are added.
The Tax Return Method. This method represents the proportions derived from the independent county populations produced by the U.S. Census Bureau.
State Estimate. The state population is estimated using the Driver License Address Change (DLAC) Method. This composite method takes account of annual births and deaths and features distinctive approaches to groups under age 18, aged 18 thru 64, and 65 and over. Driver license address changes, tax return data, Medicare enrollment, and immigration data are among the data used in this method, which includes survived cohorts of household population, group-specific migration estimates, and group quarters population.
1 Of the county methods used during the 1990’s, the Household Method was used in estimating county proportions of state population until January 1997. This method was dropped in the following year.
The Household Method estimates the annual proportionate change in each county’s distribution of households compared to the prior year. The estimated change is applied to the prior estimated distribution of the household population. Group quarters estimates are then added.
Sources. Data used in estimation models come from administrative records of several state and federal government departments and agencies, as well as numerous local jurisdictions. Since timeliness and coverage in these series vary, corrections, smoothing, and other adjustments may be applied.
Accuracy. Data and models used to produce population estimates are subject to measurement and non-measurement errors. The data and estimating models were thoroughly tested with decennial census results that provide benchmarks for the estimates series. The total state estimate was within one-half of one percent (0.5%) of the 2000 census count; county estimates varied by an average of 1.9 percent, and city estimates by an average of 5.6 percent.