ACS Data Users Group

I'm trying to get detailed data on languages spoken at home to do an LEP study. I'm able to get the data by PUMA; however, my city is lumped in with two other cities, with no way to segregate them.

Is there some other way to do this? We're on the hook for non-English language support but can't get granular-enough data to figure out which languages we have to service. Thanks in advance.

Hi,

ACS Table B16002 provides detailed language data at the household level.

Data is available for municipalities and Census Tracts. You may need to isolate the tracts in your city and then aggregate them. The ACS handbooks provide examples of the aggregation. However, if your city is defined as a census place, you may not need to rely on the tract data.

See the spreadsheets for ACS 2018-2022 posted here: https://demography.dola.colorado.gov/assets/html/acs_spreadsheets.html

for an example of the municipality-level data. (This was generated using tidycensus and openxlsx)

Cheers

AB

Thanks! I was able to pull down the table for Buena Park. However, I need people rather than households. The law we're trying to satisfy is AB 1638, which specifies "the 5% or more of the population that speaks English less than 'very well.'”

Is there a version of this table that counts noses instead of front doors?

You might try B06007: Place of Birth by Language Spoken at Home and Ability to Speak English in the United States

Thanks. All it shows is "Spanish" and "Other Languages."

This shouldn't be this difficult.

Sorry, I misunderstood your question. I don't think the ACS includes person-level detailed languages. Perhaps the law can be amended to refer to households?

That would be nice, but since it's already law, I doubt it'll happen. The Assembly probably assumed (as I did) that the Census Bureau would make this kind of data easy to get.

Perhaps whoever wrote the law should have checked what's actually available in the ACS before the law was passed.

I get difficult-to-answer questions like this a lot.

For city (sumlevel 160) and tracts (sum level 140), have you looked at table C16001 ? There're nine non-English languages listed

Grid View: Table C16001 - Census Reporter

Why just nine languages?

About 7 years back, Census execs determined they were publishing too much detail, on infrequent categories, and that this posed personal identifiability risks. So now, the only languages with published tabulations by Census are those nationally widespread enough.

(... so, for more languages, how do you feel about 2015 data??)

It looks like B16001 will do for a start. It has more languages split out.

Well, fooey. The data's almost ten years old. Back to C16001.

This has been a long-standing problem since language detail was pared back in reported ACS tables (per Todd's comment below).

Outside of the resources shared by others, you may also be able to use additional (non-ACS) sources such as school district data. You can find data for the Buena Park school district data here: www.kidsdata.org/.../table

If you're able to get the data by PUMA, you could do an allocation to the place level using Geocorr.

https://mcdc.missouri.edu/applications/geocorr.html

Table B16007 has gross language categories by age for persons over age 5. (Spanish, Other Indo-European, Asain, and Other)

You will not be able to get detailed languages from this table. If you need to drill down to specific language groups, allocating things from the PUMS is probably your best bet, per Glenn Rice

BTW, ALL of this information is available from the ACS table shells (https://www.census.gov/programs-surveys/acs/technical-documentation/table-shells.html) You may wish to become familiar with this document.

Cheers--

AB

Geocorr is upset with me, but I'll keep trying.

I wonder: how accurate will the results be? I'd thought about using a population-based adjustment, but that assumes the incidence of languages is uniform across the three cities in the PUMA. I know that isn't the case.

There is a special file that is created every 5 years. and is in the Census Redistricting Office that uses 74 Language Minority Groups to compute the percent of LEP individuals most particularly LEP individuals who are Citizens of Voting Age by State, County and Minor Civil Division, this matched Section 203 of the Voting Rights Act. There is a table that does LEP of all sorts by Spanish and Other Languages, that follows the typical summary levels. I think it is B1608. It includes Citizens and non-citizens. The p;roblems with this are mainfold. With a reule like AB 1608, is their an age range for LEP. Households are often used for other things, because one can expect those in ahousehold to care for one another. But this is a real problem. The VRA Section 203 file is created and modelled.

Andy Beveridge

Dear Lance,

I've had a chance to look at C16001 for Buena Park CA and it seems to show that the only 2 languages that would be needed to satisfy the 5% criteria in the statute are Spanish and Korean. (The 5 year + population is 78074 (ACS 2022 1-year period). That should correspond to a threshold of about 4000 non-English speakers. The "other languages" category is less than 4000 so you don't need to look for other groups.

I looked up your linkedin profile and according to your profile you have a BS in computer science. I have a computer program that will adjust PUMS data based on tract level ACS detail tables to produce tract level "synthetic" data. You can then produce a tract level table using any PUMS variables. While I would be reluctant to use the program in a legal context, the program would give you a "leg up" on what languages you might be dealing with. Go to dorerfoundation.org and using the links at the top of the page go to the "Contact Us" page. Send me an email using that address and we can communicate via email. The mission of Dorer Community Service Foundation is to provide consulting to nonprofit 501(c)(3) organizations and governments. So far all of our consulting has been pro bono. I should be able to help you out with possible approaches to your problem. I have used an earlier version of the program on a project with the Commonwealth of Virginia.

Best,

Dave Dorer

Thanks. I think I have what we need now from C16001, though not to the granularity we'd like. I hate to extrapolate from the census tracts since they seem to be a blunt instrument. I wish we could get more data about everything at the block or block group level.

Yeah, losing Table B16001 at the tract level as others have referenced has really hurt my department's ability to address our LEP obligations. The release of B16002 at the tract level with the last ACS did add back a little detail so that was a very slight glimmer of hope.

Recently, I came across Table B05006 (Place of Birth for the Foreign-Born Population in the United States) and I've wondered about using this table's information in concert with B16002 to "reverse engineer" what languages we might want to focus on when all we have to go on is a generic geographic description (Indo-European, etc). Since they're all based on samples of a 5-year timeframe, it's fraught with issues of comparison but when you're stuck between a rock & a hard place with no alternatives, why not?...

Steve

Dear Steve,

What is the city ? If you can send state name county and csd (county subdivision) or place (fips and name) or csd (name and fips) I can run it using PUMS data and my Small Area Estimation program.

Best,

Dave

I would take David up on this. The Census uses a very complex Bayesian approach to estimate languages with small numbers of people who speak them, and are Limited English Proficient. Small area estimation is the only way to do this.

Dear Steve and Andy

I assume that this is CA in which case the criteria is languages spoken by 5% or more of the population. Hopefully you won't have to "drill down" too far on the PUMS LANP variable which has over 100 different languages.

Dave

Hi David, I'm actually up in Washington State and work for Snohomish County Public Works (30 miles N of Seattle). At the urging / pressure of our federal designate (WSDOT), we've expanded our Ttile VI analysis beyond just what is required under NEPA to essentially include all or our projects so we look at everything from intersection improvement projects up to countywide programmatic projects.

BTW, I'm not the person actually responsible for the Title VI analysis. I'm a GIS Analyst tasked with supporting that role so I'm responsible for updating the data and maintaining a web map based application we use to provide access to the information.

I'd be careful about making assumptions based on place of birth. A person born in India has at least 160 languages to choose from; someone from Bosnia could speak any one of seventeen languages, including German and Italian. Even someone born in Canada could be Francophone and low-English proficient.

Hey, I freely admit that this is a flawed approach but, when you have identified a LEP population that speaks an "Indo-European Language", how can you realistically respond as required by law? There's no translation available for "Indo-European" so what do you do, because you have to do *something*.

The examples you cite are all valid but it seems to me that, in my Indo-European example, examining which Indo-European origin populations are represented in the area of interest, and then identifying which majority language they speak is a good faith attempt at satisfying the requirements of LEP under Title VI.

This is why I found and joined this site was to see how others are actually addressing this. When you're at the end of the line, we get a lot of Moses coming down the mountain with 10 Commandments telling us that we should do X & Y but it's devoid of additional guidance of "ok, you must do X & Y and here's how you accomplish it given the lack of specific information.."

Anyways, that's for voicing those concerns. I did not express it directly in my post but I have thought about that as a definite con with my idea.

Dear Steve

I'm working on an R function that creates an LEP table for a Place or CSD. The table contains all the languages for the PUMS LANP variable that occur in the PUMA. 180+ potential languages;

What is the city/town (CSD or Place) I have that you state fips is 53, It sounds like you need a county. Can you send the FIPS code. For a city/town I need the Place FIPS

or the county & CSD fips codes. If you want me to run the entire state contact me at

info@dorerfoundation.org

If you want the whole state we will need to execute a Memorandum of Understanding (MOU) laying out the ground rules. There is no charge.

The program seems to run. I'll include the function in the next release of the "PAT" package. I'm testing the function/program right now. I need to test it on a county with multiple PUMAs.

dorerfoundation.org/software.

Since you are a government you can use the program (free to government agencies and 501(c)(3) organizations)

You can us the program but be aware:

# ADDITIONAL DISCLAIMER #
# #
# This software is intended for educational, research and reference #
# purposes only. The user should use their judgment in evaluating any #
# results or conclusions produced using this software. The user is #
# advised to verify any results produced by this software by reviewing #
# other independent sources of information. #
# #
# The Dorer Community Service Foundation Inc makes no warranty that #
# this software is error-free or as to the accuracy, completeness, or #
# suitability for any use, or as to any result arising from its use. #

Dave

Dear Steve,

I have the necessary FIPS codes: State 53 county 061. There are 1771 census tracts. This will take about 2.5 days using my desktop machine.

Please email to discuss -- we can zoom if needed. I have an account.

Dave