What are Summary Files/Counts
For the Decennial Censuses, the Census Bureau has for many years conducted effectively two surveys, one 100% survey, in which every single person in the country is supposed to be accounted for, and one sample survey which is done for a random selection of households. For Census 2000 the sample survey was done for 1 in every 6 households.
One of the reasons the Census Bureau has done this is because the sample survey has many more questions than the 100% count, and it would simply be too expensive to do it for every household and is mostly unnecessary, because the sample in the sample survey is so large the numbers are very close to the 100% count anyway. The sample survey however contains a much more detailed profile of US than the 100% survey which only asks several questions such as sex, age, race.
These “surveys” are then processed into summary counts or files.
For Census 2000, the Census Bureau has produced four Summary Files (SF). The first two (SF 1/2) are based on the 100% sample and the later two are based on the sample survey (SF 3/4).
So technically speaking, if the data you are looking for is located in the first two summary files (SF1/2), you should use those numbers, because they are more “accurate”, otherwise you have to use SF 3/4.
What is in SF2/4?
For Census 1990 the Census Bureau has produced four Summary Tape Files (STF). They are equivalent to the Summary Files in Census 2000. Summary Tape File 1/2 (STF 1/2) contains 100% data on age, race, sex, marital status, Hispanic origin, household type and relationship, occupancy/vacancy status, tenure, units in structure, contract rent, meals included in rent, value, and number of rooms. Summary Tape File 3/4 (STF 3/4) contains sample data weighted to represent total population. Again as in Census 2000, STF 2/4 data tables are iterated by detailed race and ancestry categories.