NDC’s Estimated 2020 Census Data
With the official 2020 Census data delayed until August, many jurisdictions are asking NDC for our “best estimate” of what the 2020 Census data are likely to show.
NDC initially tried to simply use 2015-2019 American Community Survey population estimates, but at the local level we found there was too much statistical “noise” in the data: numerous Census Tracts showed population decreases simply from statistical ’rounding’ in the data. So NDC began a process of estimating 2020 Census Block populations using the American Community Survey data to identify where population growth since 2010 has occurred.
A caution: while estimated data can be very useful, they remain estimates. The only thing guaranteed about all estimates is that they will be wrong. The question is how close can we make them, even though we know they will not be exactly correct.
Only use this data if you understand that all estimated data are always wrong to some degree, and it is impossible to know that degree of error until the official 2020 Census data are available.
How NDC Built this 2020 Estimated Total Population Database
Here is the process NDC used to develop this data:
|Step||Selection||Action||# geographic units|
|1||2010 Census Block Group population below 250 and 2015-2019 ACS population below 250||All Census blocks keep 2010 Census count||124 block groups|
|2||Block Groups with 100% population loss compared to 2010 Census Population not included in Step 1||All Census blocks keep 2010 Census count||1 block group|
|3||Block Groups with 0 population in 2010 Census but >250 population in 2015-2019 ACS||Divide 2015-2019 ACS by area to each block within the block group||1 block group with two Census Blocks|
|4||Tracts with <25 Housing Units and <20% living in group quarters according to 2015-2019 ACS, not included in Steps 1 or 2||All Census blocks keep 2010 Census count||0 block groups (step is included for the record)|
|5||2015-2019 ACS Population Group Quarters > 20% and 2010 Census – 2015-2019 ACS growth rate within one standard deviation||All Census blocks keep 2010 Census count||177 block groups|
|6||Tracts with >20% of housing constructed after 2010 and % in group quarters under 20%||Applied the higher of:
(a) 2010 Census to 2015-2019 ACS Block Group growth rate
(b) 2006-10 ACS to 2015-2019 ACS Block Group growth rate
to all Census Blocks in each Block Group
|348 block groups|
|7||Remaining Block Groups||Applied 2015-2019 ACS growth rate to 2010 Census total population counts||Approx. 22,540 block groups|
|8||Prison Population Edits||Removed population in California State Prisons (see notes below).|
 High ACS margins of error in small units of geography often lead to populations of zero where Census 2010 counts small populations and wide population swings in small-population block group. These swings result from statistical noise in the ACS methodology, rather than actual population change, so we remove them.
 About 30 coastal water areas and 90 mainly rural block groups
 This must be an error in ACS; the block group is the UCSD campus, including dorms, in San Diego.
 Year of household construction and group quarters population are only available at the tract level in the American Community Survey.
 Because of how ACS measured group quarter populations, using growth rates based on ACS data are inaccurate.
 Because of how ACS changes year-over-year, only using one growth rate may lead to an irrational conclusion of population loss. Using one growth rate rather than the higher of the two would lead to an estimated state population of only approx. 35 million. This method corrects for this problem.
California state law directs the California Statewide Database to zero-out state prison populations and to county state prisoners at their last known address. The state is not releasing any data related to this work until they release the official 2021 Redistricting Dataset. Without that data from the state, NDC cannot add the prisoners to the Census Blocks containing their last known address. But we can, to the best degree possible given the data available, zero-out the state prison populations.
NDC asked the California Statewide Database staff whether they will zero-out federal prison populations, and the Statewide Database staff responded that the decision on that has not yet been made, so NDC has not made changes to Census Blocks containing Federal prisons.
Here is the list of state prisons NDC has removed from the population counts:
Prisons Excluded from count:
|Avenal State Prison (ASP)||Avenal||CA||93204|
|California State Prison, Los Angeles County||Lancaster||CA||93536|
|California City Correctional Facility||California City||CA||93505|
|California Correctional Center||Susanville||CA||96127|
|California Correctional Institution||Tehachapi||CA||93561|
|California Health Care Facility, Stockton||Stockton||CA||95215|
|California Institution for Men||Chino||CA||91710|
|California Institution for Women||Corona||CA||92880|
|California Medical Facility||Vacaville||CA||95696|
|California Men’s Colony||San Luis Obispo||CA||93409|
|California Rehabilitation Center||Norco||CA||92860|
|California State Prison, Corcoran||Corcoran||CA||93212|
|California State Prison, Sacramento||Folsom||CA||95671|
|California State Prison, Solano||Vacaville||CA||95696|
|Calipatria State Prison||Calipatria||CA||92233|
|Centinela State Prison||Imperial||CA||92251|
|Central California Women’s Facility||Chowchilla||CA||93610|
|Chuckawalla Valley State Prison||Blythe||CA||92225|
|Correctional Training Facility||Soledad||CA||93960|
|Deuel Vocational Institution||Tracy||CA||95376|
|Folsom State Prison||Folsom||CA||95671|
|High Desert State Prison||Susanville||CA||96127|
|Ironwood State Prison||Blythe||CA||92225|
|Kern Valley State Prison||Delano||CA||93216|
|Mule Creek State Prison||Ione||CA||95640|
|North Kern State Prison||Delano||CA||93215|
|Pelican Bay State Prison||Crescent City||CA||95531|
|Pleasant Valley State Prison||Coalinga||CA||93210|
|R.J. Donovan Correctional Facility at Rock Mountain||San Diego||CA||92179|
|Salinas Valley State Prison||Soledad||CA||93960|
|San Quentin State Prison||San Quentin||CA||94964|
|Sierra Conservation Center||Jamestown||CA||95327|
|Substance Abuse Treatment Facility and State Prison, Corcoran||Corcoran||CA||93212|
|Valley State Prison||Chowchilla||CA||93610|
|Wasco State Prison||Wasco||CA||93280|
- The list is based on California Department of Corrections Population Reports. It is unknown exactly what the State will determine regarding Federal prisons, County jail populations, and juvenile prison populations, so we continue to include them in the database.
- In all but three cases, we have zeroed out the entire population of the block containing the prison. While this metric might inadvertently include a non-incarcerated population (e.g. guards), the American Community Survey data indicate that at or near 100% of the population of the block is in “Group Quarters”.
- In three cases (the California Institution for Men in Chino, the California Institution for Women in Corona, and the Deuel Vocational Institution in Tracy), it was determined that these blocks may include substantial non-incarcerated populations. We therefore estimated the proportion of the population in prison using the Group Quarters metrics from the American Community Survey 2015-2019 and removed that proportion of the population from the block’s population.
Converting to 2020 Census Block Geography
The Census Bureau’s file is far from perfect, so some oddities in the data may result from this conversion. Again, if you are unfamiliar with this process and the errors that can result, you should probably not be using this data.
If you do not understand that this process introduces unavoidable error, and if you do not understand how to work with data even understanding there are errors in it, you should not use this data.
Citizen Voting Age Population Data
The file includes estimated Citizen Voting Age Population (CVAP) data from the Census Bureau’s Special Tabulation of American Community Survey Data. The CVAP data are disaggregated from the Census Bureau’s Block Group-level data to individual Census Blocks based on each Block’s share of the Block Group’s population in each separate ethnic and racial category. The disaggregation was done using 2010 Census geography, then converted to 2020 geography using the 2010 to 2020 equivalency file mentioned above.
Caution: these data are estimates of estimates, and thus involve unavoidably large potential errors at the local and individual Census Block level.
Ready for data?
These data are offered as a public service with no guarantee, assurances, or promise about the accuracy of the data provided. The only guarantee is that these data will be incorrect, though we hope (but do not promise) they will be off by only small amounts.
The 2020 Census was a uniquely challenging project, and neither NDC nor anyone else can precisely predict how the official 2020 Census data will look once it is released.
Again: if you do not understand that all population estimates involve errors, and if you are not comfortable using with data knowing there are unavoidable estimation errors in it, you should not use this data.
By clicking the following link to the *.zip file of the data (provided in a Maptitude Geographic File format), you affirm “I understand the imprecision of all population estimates and affirm that I will not hold NDC responsible in any way for the accuracy of this data or of any work that I perform using this data.” (Link fixed 5/22/2021.)