Census Data is collected every 10 years by mail surveys to every household with primary data collection fields of population, gender, race and number of occupants. Overcounts are known to exist in student and dual-residence populations, while undercounts are a given for homeless and illegal immigrants. Errors arise from both sampling and non-sampling methods. Sampling errors are misrepresentations of a larger population based on the selection of the representative sample. Accuracy of the estimates depend in part on how typical the people within the sampled housing units are to others within their area. The Census estimates sampling error based on the population and its extent of variability. The long-form of the Census is distributed to a subset of households, and includes additional data related to households, economy, employment, occupation, as well as occupancy and building structure data. Household fields include: Total Households, Total Families, Households by Household Type by Household Size, Persons in Group Quarters by Detailed Group Quarters Type, Persons 25 Years and Over by Educational Attainment, Persons 16 Years and Over by Labor Force and Employment Status, Median Household Income, Median Family Income, Per Capita Income, Persons by Poverty Status in Previous Year. Occupancy fields include: Total Housing Units, Housing Units by Occupancy Status, Vacant Housing Units by Vacancy Status, Occupied Housing Units by Tenure, Median Housing Value, Median Contract Rent. The 2000 census marked the last year that the long form was distributed to 1 in 6 households (a 17% sample), and in 2010 the actual sample size for the long form was 2.5% per year (totaling 12.5% over a 5 year cycle). In 2000, one household answered the long form questions on behalf of 5 other households, and in 2010 a single household provides data for between seven other households (5 year averages) and 39 other households (1 year averages for larger places).