skip to content
Primary navigation

Technical Notes

View: BDCD Tool Home | User's Guide | Technical Notes | Examples | LMI Help

Sample Size and Margins of Errors

*Please refer to the “American Community Survey” site for a full technical description and presentation of sampling margins of error recommended by the Census Bureau”.*

Bachelor Degree and Career Destination (BDCD) data tool was developed by using the American Community Survey (ACS) of the United States Census Bureau for 2012-2016 1-year sampled population. In a survey sample, there is always room for error. The American Community Survey provides data on self-reported residents of Minnesota who earned a Bachelor’s degree in relation to the U.S. population as a whole. The ACS accounts for variations and error within their data collection before it is released and use by the public. Given the accumulation of data, the variation of sampling is accounted for, given a certain degree of significance for each year of sampling that was used in the BDCD tool.

Where the error might arise in the tool

The BDCD tool focuses on occupations, median income, and labor force indicators for bachelor degree holders by their fields of study. Given these factors, the complete sample for Minnesota residents with Bachelor’s degree gets fragmented by fields of study, and further by occupations within and across each field of study. The percentage frequency distribution is hence dependent on the number of degree holders within the field of study and their occupational choices and other characteristics that may be hard to account for. Some fields of study are obviously more common than others. An example would be Business, which would have more participants than say Library Sciences. Given this, the margins of error for the sample used to find the percentage frequency distribution of occupational choices for those who had Business as their field of study would be smaller than those who opted for Library Science. In this case the data for business degree holders could possibly be more significant, making the data for Business degree holders more reliably representative of Business degree holders than for Library Sciences.

ACS Data Collection

The methods used to obtain the American Community Survey data collection were internet, mailout/mailback, computer assisted telephone interview and computer assisted personal interview. All responses for data collection were accepted within a three month period, September through December. According to the U.S. Census Bureau, the ACS engages in a two phase, two stage sample design to address random sampling for each method used for data collection (States, 2019). The survey address Group Quarters and Housing Unites to cover the entire population of the U.S. and Puerto Rico. Housing units consists of addresses in all county and county equivalents in all 50 States, including the District of Columbia. Certain groups of individuals are excluded from the survey sampling based on weighted estimations. These individuals are seen as fragment residents of the populations they are adjacent to. An example would be citizens who are associated with soup kitchens or domestic violence shelters.

ACS Weighted Data

Weighting a survey sample is important in assuring that the data reflects the whole population. The American Community Survey was weighted by two measures, each sampled person recorded and a weight of each sampled housing unit recorded. A person’s characteristics determined the estimations of a person weight by tabulating the person, households, families or housing units. To determine household weight the individuals housing unit weights are assigned by exactly one weight to reflect estimates of characteristics. If personality respondent’s individual characteristic weight is 40 then their household weight would also be 40. The data collected for sampling in the BDCD tool was measured for a five year weighted period which accounts for 60 months of collected data pooled together. The weighted data over a five year span uses the same method as a one year weight as described above, with a few adjustments. The modifications concerned geography, month-specific weighting steps, population and housing unit controls.

ACS Significance of Error

Sampling errors are developed for the ACS based on the sample surveyed and the corresponding value that would be obtained if the entire population were surveyed based on Census data. The American Community Survey yields are approximations of the actual numbers that would be obtained by interviewing the entire population. To account for error in the population given different characteristics, the Census Bureau estimates the specific chosen sample results then compares the margin of error from one sample to another.

The ACS design for finding margins of error were independently selected using three levels of significance to determine the variations or error within sampled population. The measures of significance examined for all ACS published margins of error are based on a 90% confidence level, meaning the data set accounts for 5% pulse or minus error margin (States, 2019) . All data from the American Community Survey were represented by the population as a whole given sample size surveyed.

Calculation used to find standard error and confidence bound:

  • Margin of Error / 1.645= Standard Error
  • Estimate - Margin of Error Upper Confidence Bound = Estimate + Margin of Error Lower = Confidence Bound

BDCD tools Reason to account for ACS Sampling Error

The information presented in the Bachelor Degree and Career Destination (BDCD) tool is subject to margins of error due to data being extracted from the American Community Survey (ACS). Sampling errors occur due to differences in sizes of the population belonging to various fields of studies. The sample margin of error measures the maximum amount by which the sample statistics are expected to differ from characteristics of the actual population. There are many factors that can cause margins of error within samples, but mainly due to the different descriptive characteristics of individuals involved. One of the biggest factors causing margins of error is the sample size within a degree field. This impacts the significance i.e. the reliability of the sample in so far as it can be expected to represent the population for a certain field of degree in Minnesota as a whole.


States, U. (2019, February 8). American Community Survey Multiyear Accuracy of the Data. Retrieved from United States Census Bureau.

back to top