Suppressed data in the 2022 5-year estimates that have never been suppressed before

tpreston 9 months ago

We have noticed there are quite a few -666666666 values in the 2018-2022 5-year ACS estimates. These are for variables that have always had values previously. For example, B19067_001 (cash public assistance income). Out of the 31,895 CDPs we are interested in 11,455 (36%) of them have a -666666666 value. Similarity, county subdivisions are at 32%, Indigenous areas are at 34%, and tracts are at 17%. We have also noticed fairly large amount of suppressed data for B19066_001 (total supplemental security income), B25064_001 (median gross rent), B25088_002 (median monthly mortgage cost), as well as others. I'm curious as to why we are just now seeing large numbers of suppressed data. I realize -666666666 means the following...

The estimate could not be computed because there were an insufficient number of sample observations. For a ratio of medians estimate, one or both of the median estimates falls in the lowest interval or highest interval of an open-ended distribution. The estimate could not be computed because there were an insufficient number of sample observations. For a ratio of medians estimate, one or both of the median estimates falls in the lowest interval or highest interval of an open-ended distribution. For a 5-year median estimate, the margin of error associated with a median was larger than the median itself.

I'm uncertain as to what changed with the 2022 5-year release that caused the -666666666 values to be published in place of estimates (all of these variables have had actual estimates in previous years). I have read through the Data Suppression information on the CB webpage, but this hasn't been updated since 2016 and doesn't provide any information on the recent change. If anyone knows what might be going on here, we would greatly appreciate any guidance or resources you might be able to provide.

Vicki Mack 9 months ago

Hi, I have sent your question to the subject matter experts. They are looking into this for you. I will let you know their response when I hear back. Vicki
Cancel
Up 0 Down

Reply

Cancel
tpreston 9 months ago in reply to Vicki Mack

Thank you Vicki Mack! We really appreciate it.
Cancel
Up 0 Down

Reply

Cancel

Vicki Mack 9 months ago in reply to tpreston

Hi Preston, I received a reply from the subject matter experts. They wrote: "

The suppressions for B19067 2022 5-year and 2021 5-year data are summarized in the table below for the two geographies.

Data Year	Sum- level	Geography	Suppressed Lines	Suppressed + zero Lines	Total Lines	% Suppressed	% Suppressed + Zero
DY2021	060	County Subdivision	10,729	22,698	36,567	29.34%	62.07%
DY2022	060	County Subdivision	11,001	22,850	36,529	30.12%	62.55%
DY2021	160	Place	10,131	21,392	31,908	31.75%	67.04%
DY2022	160	Place	10,304	21,827	32,186	32.01%	67.82%

I included zero lines as well as lines that were explicitly suppressed. The data user appears to be accessing ACS data via the API which means that both zeros and these suppressed estimates will appear as "-666666666".

My understanding is that summary level 160 includes a combination of incorporated places and census designated places. All the suppressions applied to this table are due to the unweighted rule of 3. As displayed above, there was a slight increase in suppressions between 2021 5-year and 2022 5-year data for B19067 at these two geographies. I am not aware of any new suppression rules that were applied to this table or to these geographies in 2022.

Checking the suppressed lines, nothing significant stands out in the data that would suggest any type of issue. It is possible that the increase in unweighted suppressions at lower-level geographies can to some extent be attributed to the decrease in ACS response rates across the two years."