ACS Data Users Group

 View Only

  • 1.  0 values in ACS margins of error

    Posted 05-05-2026 10:25 AM
    Edited by Jonathan Nigrine 05-05-2026 10:27 AM
    I have a question about handling standard errors of ratios when the denominator is 0. The Census publication Instructions for Applying Statistical Testing to American Community Survey Data provides instructions for cases with numerators and denominators, but not when the denominator is 0 - even though both 0 values often include margins of error in ACS data. At the bottom of page three, there is a formula for the SE of a proportion or percentage that includes the number of cases in the denominator. That will obviously generate an error where N = 0.
     
    An example: in Clio, MI, 2016, table B18101B, Age by Disability Status (Black or African American Alone), the values for the number under 18 years (B18101B_002) and for the number with a disability (B18101_003) are both 0 +/- 10.  A 0 in the formula's denominator obviously won't work, but is there any way to handle this? I haven't found any in the Census documentation.
    Thanks!



    ------------------------------
    Jon Nigrine
    Greater Flint Health Coalition
    ------------------------------



  • 2.  RE: 0 values in ACS margins of error

    Posted 05-05-2026 10:50 AM

    Jon, I don't think there is official Census Bureau guidance for this exact case. The published ACS formulas define the proportion/ratio as A/B, so when the denominator estimate is 0, the derived estimate is undefined and the MOE formula doesn't apply.

    In practice, I would flag that cell as N/A or "not calculable," rather than report 0 or try to compute an MOE. 

    If this measure is needed, the best option may be to aggregate to a larger geography where the denominator is nonzero. 

    A similar issue came up in an earlier ACS DUG community post that might be helpful: link



    ------------------------------
    Mark Mather
    Associate VP
    PRB
    ------------------------------



Related Content