Several methodologies may be employed to estimate the Hispanic undercount and hence a correction for the 2000 census. The following documents the methodologies used in this model.
Doug Bachtel with the University of Georgia estimates the Hispanic population in Hall County was undercounted in the 2000 census by a factor of 2.5. When questioned he references Beatrice F. Piddock with the Census office in Atlanta. When contacted Ms. Piddock stated that Census statisticians were not content with the accuracy of their estimate of the undercount.
An apartment owner in Hall County surveyed all Hispanic renters in an apartment complex immediately after the 2000 census and noted that only half of the renters had responded to the census. This corresponds to a correction factor for the Hispanic population of 2.0.
Hall County Hispanic births and deaths are collected by the Georgia Division of Public Health. Assuming local Hispanic culture and population structure approximates that of their native country then the birth and death counts may be used to calculate the Hispanic population from the respective native country rates.
The CIA World Factbook 2000 estimates Mexico's 2000 crude birth rate at 23.15 births per 1,000 population. The U.S. Census Bureau's International Data Base estimates Mexico's fertile women (aged 15 to 44) to be 25,020,616 which combined with the CIA data yields 9.3 children born per 100 women aged 15 to 44. This number is referred to as the birth rate.
Assuming the age and gender structure of Hispanics in Hall County as reported in the 2000 census is correct, a population pyramid can be constructed.
From the 9.3% birth rate and the 1,031 Hispanic resident births in 2000 an estimate of the number of women aged 15 to 44 can be calculated. That estimate combined with the population pyramid data yields an Hispanic undercount correction of 1.77.
This study does not attempt to estimate the total Hispanic population from Hispanic death counts because the local Hispanic death rate is so low. The birth to death ratio for Hispanics in Hall County in 2000 is 54.3:1 while that for non-Hispanics is 1.9:1. One factor evidenced by this ratio is that Hispanics who immigrated to Hall County came to labor, with the majority arriving over the past 13 years. Hence, there is a preponderance of healthy adults. Another factor is that it is not uncommon for an older, first generation Hispanic immigrant to return to their native country to die.
While Census Bureau does not endorse an estimate of the 2000 undercount their official estimate of the 1990 undercount is 2,257. If we assume that this undercount is composed primarily of Hispanics then the 1990 correction factor for the Hispanic count is 1.5. It is not unreasonable to use this as another estimate of the correction factor for the 2000 Hispanic undercount.
Another factor that can be used to estimate the number of immigrants to Hall County is the average number of residents per housing unit and the net increase of housing units constructed. The background culture's (fully assimilated multi-ethnic population) number of residents per housing unit has declined steadily from 3.59 in 1960 to 2.7 in 1990. This decline reflects the decline in the background culture's optimum family size. Census survey data can be used to approximate the number of residents per housing unit for Hispanics. The Census 2000 March Supplement lists 4.5 and 4.8 persons per household for households of Mexican origin living in the United States. The 2000 census lists the total number of housing units at 51,046, an increase of 12,731 units from 1990. However, building permits in Hall County for the same period yield an actual increase of over 13,900. Therefore, census missed 1,169 residential units. This difference between actual housing units and reported housing units represents an indirect measure of the Hispanic undercount of the 2000 census. Using 4.65 persons per household (the average between 4.5 and 4.8) 1,169 x 4.65 or 5,436 Hispanics were missed by the 2000 census.
Yet another approach is to analyse the population and housing increase over the decade. Hall County non-Hispanic population increased from 1990 to 2000 by 21,165 individuals. Using the 1990 residents per housing unit number of 2.7, we calculate that the additional non-Hispanic population of 21,165 individuals required approximately 7,839 additional housing units. That leaves 6,061 residential units for the increased Hispanic population. At 4.65 persons per household unit, that is sufficient housing for 28,184 Hispanics. The 2000 census records 20,427 additional Hispanics in Hall County for that decade. By this calculation, the estimate of the Hispanic undercount is 7,757.
Averaging these two estimates yields a correction factor of approximately 1.25+.
Though housing unit removals are not handled in this data,
building permit data is still being collected from some municipalities, hence the '+' sign.
The above estimates of the Hispanic undercount correction factor range from a low of 1.25+ to a high of 2.5.
The 2.5 estimate has no backing data, therefore will not be considered by this study.
The 2.0 estimate based on a survey at an apartment complex is potentially credible but is not random. Hence, it is possible for a few strong voices within the complex to have dominated the response of the occupants to the 2000 census. Still, it is a data point that should be considered. The strongest counter point to a 2.0 undercount estimate is that two other measures, birth rate and residents per household, are pushed hard outside the cultural norm.
The 1.77 correction factor is based on a dual linkage of assumptions. The first being that the birth rate of local Hispanics corresponds to that found in Mexico. This is a weak assumption. In 1970 the birth rate in Mexico was the highest in the world outside of Africa. Most women were not educated, had no employment opportunities, did not have access to contraceptives, and were at least nominally Catholic which does not technically allow the use of contraceptives. Due to a strong government program the birth rate has been brought down sharply. However, it could be argued that a large variation exists in Mexico and the specific cultural roots of our Hispanic immigrants has not been studied. Several similar arguments could be made for either higher or lower local Hispanic birth rates than are found in Mexico. Even the population pyramid data is suspect because it would be easier for an undocumented Hispanic to be a housewife than a male laborer.
The 1.5 correction factor based on the 1990 estimate is probably a good rough estimate because the pattern of Hispanic reticence regarding the 2000 decennial census is likely to be constant for first generation Hispanics.
The 1.25+ correction factor estimate is based on incomplete data and is ignored.