Smoothed County Population Estimates

for 1991-1998

Methodology

County population counts for April 2000 from the 2000 Census were released by the Census Bureau in March of 2001. Smoothed estimates of the population of North Carolina Counties for July 1, 1991, 1992, 1993, 1994, 1995, 1996, 1997, and 1998 were released in May of 2001 by the State Demographer. These estimates were based upon a process of smoothing the revised estimates released in June of 2000 to more nearly match the trendline between the values for April 1, 1990 and April 1, 2000 from the 1990 and 2000 Censuses. This smoothing process was redone in early 2008 to account for 2000 Census corrections received between May of 2001 and November of 2007. The resulting new estimates were released by the State Demographer in April of 2008.

For the revised estimates of population for North Carolina and its counties for 1991-1998 which were released in June of 2000, the county boundaries are those in effect for the 1990 federal census, except for a series of small revisions for 27 counties. These revisions are based on some actual boundary shifts, as well as data corrections. We released the new April 1, 1990 county totals in May of 2000.

Basic Technique, Revised Estimates

In March of 2000, the Population Division of the United States Bureau of the Census released their revised 1991 through 1998 county population estimates for North Carolina. They used an administrative records technique similar to the 1980's technique of the same name. First, we revised their new estimates for some corrections to institutional populations for April 1990 and for some of the estimate dates. Then, we used the revised state totals produced by this process and the same institutional population corrections to revise our corresponding estimates based on a ratio/correlation technique. Finally, we averaged the two sets of revised estimates for each year.

Ratio/Correlation Estimates

The same basic technique was used for each of the eight sets of July 1 estimates. First, we used changes in the numbers of MEDICARE enrollees to estimate the population 65 years of age and older. MEDICARE enrollee counts for July 1, 1989 and July 1, 1990 were used to estimate values for April 1, 1990. The latter were subtracted from values for July 1 of the estimate year and the resulting differences were assumed to approximate the changes in the numbers of person aged 65 and older during the estimate period (from April 1, 1990 to July 1 of the estimate year). These changes were added to the April 1, 1990 population counts for persons 65 years of age and older to create the final estimates for this age group.

Second, we used a standard ratio/correlation method to estimate the population aged 0-64 for all counties except Pitt, which had a countywide special census in April of 1998. The data series used were automobile and truck registrations (X1), school enrollment in grades 1 through 8 (X2), and a three year sum of births (X3). For the estimate year, the three year sum of births was the sum of calendar year values for that year and the two prior years; for 1990, the sum of values for 1988, 1989, and 1990. In order to remove the effects of United States military deployments in the Persian Gulf in late 1990 and early 1991, automobile and truck registration totals for the base point (January 1991) were estimated as the average of corresponding values for January 1990 and for January 1992. The prediction equation for each North Carolina county, excluding Pitt, is given by

y = -0.038424 + 0.455019 * X1 + 0.320879 * X2 + 0.248463 * X3,

where y represents the estimated ratio of percentage shares of nongroup quarters population and each of the series indicators (X1, X2, and X3) represents the ratio of percentage shares of the associated variable. The equation coefficients were derived by least squares regression, using series indicator and population values for 1980 and 1990. The results of this equation were adjusted for the extra 1/4 year of the estimate period by linearly expanding the change in y from the assumed 1990 value of 1.0. The nongroup quarters 0-64 population estimates derived from this equation were combined with independent estimates of the population of military barracks, college dormitories, and other institutions to yield estimates of the total 0-64 population.

Finally, we added a value for Pitt County for the 0-64 population for the estimate date based on the special census. For the 1998 value, growth from the 1990 census to the 1998 special census was extrapolated forward to July 1, 1998. In calculating this value for years between 1990 and 1998, a provisional value for each year was first calculated without the special census, using slightly different ratio-correlation coefficients based on 100 instead of 99 counties. Then, the ratio of the true value to the provisional value was assumed to change geometrically with time from 1.0 on April 1, 1990 to the ratio of the extrapolated value to the provisional value on July 1, 1998.

Subtraction of the sum of the county populations 65 and older for each year from the corresponding state total population estimate produced a control for the 0-64 population estimates. There was no state control for the population 65 and older.

Methodology --- Age, Race, and Sex Detail

To calculate the populations for the different age, race, and sex groups, I used a relatively simple process. First, April 1, 1990 and April 1, 2000 Census counts for each county by 384 age, race, and sex cells (96 age values: 0,1,2,...,94, and 95+; white/other; male/female) were interpolated (age cohort interpolation for non-institutional populations; age group nterpolation for institutional populations) to each of eight (8) dates (July 1 of each year from 1991 through 1998). (In early 2007 and again in early 2008, this process was revised.) Next, the 384 cells for each county for each year were proportionately adjusted to sum to the county total population for that year. Finally, the appropriate age cells were summed to form each group.

Last Update: April 9, 2008