Supplement: Analysis of the 2023 Expert Report
The previously sealed “Expert Report” for the 2023 Arizona case of Mi Familia Vota v. Fontes presents “estimates” which “suggest” that “slightly less than one percent of noncitizens” are registered to vote. As documented below, all of these estimates are lowballs.
Pages 65–68 of the report calculate the frequency of non-citizen voter registration in Arizona from voter databases and Arizona Department of Transportation (ADOT) records.
The report claims to “present minimum and maximum estimates” of non-citizen voter registration and explains that the “maximum estimates include all individuals whom the ADOT database did not indicate were citizens at the time of registration.”
However, the maximum estimates were omitted from the report, and the table of results has blank cells on the lower right where those figures were excluded:
Based on data in the table, Just Facts calculated that the missing maximum estimate for the portion of non-citizens registered to vote is 56%. The author of the report confirmed this is correct in an email to Just Facts.
Pages 68–70 of the report calculate the frequency of non-citizen voter registration in Sedgwick County, Kansas from the voter registration records of people who recently become U.S. citizens.
This analysis produces lowball estimates because it only examines a narrow category of immigrants who are unlikely to commit voter fraud.
As confirmed by the author of the report, the only people included in the analysis were a subset of legal immigrants who became citizens since 2016. All other non-citizens, including unauthorized immigrants, were excluded.
Immigrants who become citizens must be in the U.S. legally, are heavily vetted for criminality, tend to be highly educated, and are among the most law-abiding groups of people in the nation. Moreover, they could lose their legal status by illegally registering to vote.
In contrast, unauthorized immigrants live in the U.S. illegally, tend to be poorly educated, and routinely commit identity fraud.
In short, the analysis is a classic case of overgeneralization in which narrow data was used to draw broad conclusions.
Pages 70–71 of the report calculate the frequency of non-citizen voter registration among people who lived in North Carolina during 2014, had a driver’s license, and filed paperwork under the Deferred Action for Childhood Arrivals (DACA) program.
DACA refers to President Obama’s decision to give effective amnesty to “Dreamers”—or unauthorized immigrants who arrived in the U.S. before the age of 16.
This is another case of overgeneralization. Given that people who apply for this program are subject to criminal background checks and must be under the age of 16 when they arrived in the U.S., it’s unlikely that their rate of illegal voter registration is similar to the entire pool of non-citizens. This is especially true of those who illegally entered the U.S. as adults. Again, these immigrants tend to be poorly educated and routinely commit identity fraud.
The report obliquely acknowledges the futility of the analysis by stating that “there is no way to know how DACA registration rates compare to other noncitizens.”
Pages 72–76 of the report calculate the nationwide frequency of non-citizen voter registration from the 2022 Cooperative Election Study, a survey of 60,000 people by Harvard University and YouGov. This is the same survey used in the 2014 Electoral Studies paper and Just Facts’ studies of non-citizen voting and registration.
The 2023 report explains that “more accurate assessments” of non-citizen registration “are now possible” due to the “important advance” of a second citizenship question in the survey. This enhancement “dramatically reduces the risk” that survey respondents who are citizens might mistakenly identify themselves as non-citizens.
However, the report also documents that the additional question had no discernible impact on the results: “Across more than 164 thousand responses” from 2019 to 2022, “there were no instances in which someone indicated that that they held a United States citizenship on one question and that they did not hold United States citizenship on the other question.”
Thus, the main benefit of the second question isn’t greater accuracy but strong evidence that scholarly critics of the 2014 Electoral Studies paper were wrong to allege that non-citizens in the survey who had registered and voted were citizens who mistakenly identified themselves as non-citizens.
Yet, the 2023 report calculates that slightly less than 1% of non-citizens were registered to vote in 2022, while the 2014 Electoral Studies paper calculated that the “best” estimate for this figure was 25% in 2008. What explains this massive difference? Changes in methodology.
First, the 2023 report excludes the following groups of people from the count of registered voters:
Also, the 2023 report doesn’t weight the survey results to make them representative of the nation’s non-citizen population. This methodology may understate voting by non-citizens because:
Finally, the 2023 report assumes that the appropriate denominator for calculating the registration rate of non-citizens matched to a voter registration database is the number of non-citizens who participated in the survey. This assumption may understate voting by non-citizens because:
[1] “Expert Report of Jesse T. Richman in the Case of Mi Familia Vota v. Fontes.” October 13, 2023. <www.justfacts.com>
Page 72:
Each year the CES [Cooperative Election Study (CES) survey (formerly known as the Cooperative Congressional Election Study or CCES)] asks respondents their citizenship status with at least one question (and more recently with two questions) and asks respondents whether they were registered to vote or not. In addition, the CES long retained the voter file firm Catalist to match survey respondents with voter-file and commercial records in an effort to verify respondents’ statements of whether they were registered to vote. In 2022 the CES switched its voter file matching from Catalist to instead use the TargetSmart database.
[2] Calculated with data from the paper: “Do Non-Citizens Vote in U.S. Elections?” By Jesse T. Richman, Gulshan A. Chattha, and David C. Earnest. Electoral Studies, December 2014. Pages 149–157. <www.sciencedirect.com>
Page 150: “Of 339 non-citizens identified in the 2008 survey, Catalist matched 140 to a commercial (e.g. credit card) and/or voter database.”
CALCULATION: 140 / 339 = 41%
[3] Email from Just Facts to Dr. Jesse Richman, April 16, 2024:
What portion of the people who identified themselves as non-citizens in the 2022 CES were matched in the TargetSmart database?
Email from Dr. Jesse Richman to Just Facts, April 16, 2024:
It was 7.21 percent. … Unlike Catalist (especially in the earlier years of the CCES) targetsmart doesn’t seem to be matching on anything other than the voter registration list. Thus, if someone is matched by them, it’s mostly because they are registered to vote.
[4] Report: “Beyond the Core and Periphery: A New Look at Voter Participation Across Elections.” By Stephen Ansolabehere and Brian Schaffner. Harvard/YouGov Cooperative Congressional Election Study, November 30, 2015. <cces.gov.harvard.edu>
Pages 3–4:
The second source of data we analyze in this paper is a 1% sample (N=2,969,951) of American adults from Catalist. Catalist is a prominent voter file firm typically providing data to progressive clients, but also to academic institutions under a special subscription plan. Catalist maintains a database of information for nearly every American adult. The database is first built on voter file records acquired from all 50 states, and then a variety of different information is appended to that data from marketing firms, census data, and other sources. Catalist also builds several of its own models into the dataset to distinguish, for example, the partisanship, ideology, and political activism of each individual. The Catalist dataset is ideal for studying voter turnout because it includes precise information on each individual’s turnout behavior and is sufficiently large so that even relatively uncommon voter types can still be studied with precision.
[5] Webpage: “Catalist Data.” Catalist. Accessed November 15, 2016 at <bit.ly>
Catalist provides data and data-related services to progressive organizations to help them better identify, understand, and communicate with the people they need to enhance, persuade, and mobilize.
COMPREHENSIVE
Our national database contains more than 240 million unique voting-age individuals. We start with a rich, reliable foundation of 185 million registered voters – a unified national voter file – collected from Election Officials in all 50 states and the District of Columbia.
Commercial data adds another 55 million unregistered individuals and enhances the entire database with hundreds of fields, including household attributes, purchasing and investment profiles, donation behavior, occupational information, recreational interests, and engagement with civic and community groups.
We also integrate data from the Census, specialty data, and media market geographies. We have records going back more than a decade, giving our clients the unique ability to see both short-term and long-term trends.
ACCURATE
Catalist data is continually updated and audited by data quality experts. Catalist’s Data Operations team adheres to a rigorous multi-step quality assurance process that ensures that each update improves the quality of the database.
THE CATALIST NATIONAL DATABASE BY THE NUMBERS:
240 Million+ unique voting-age individuals
185 Million+ registered voters
55 Million+ unregistered voting age persons
NOTE: 240 million voting-age individuals = 96.3% of the U.S. adult population of 249,291,898 people in 2016 (when this Catalist webpage was accessed). [Calculated with the dataset: “Annual Estimates of the Resident Population by Single Year of Age and Sex for the United States: April 1, 2010 to July 1, 2019.” U.S. Census Bureau. Accessed May 9, 2024 at <www.census.gov>]
[6] “Guide to the 2022 Cooperative Election Study.” By Stephen Ansolabehere, Brian Schaffner, and Marissa Shih. Harvard/YouGov Cooperative Election Study, August 2023. <dataverse.harvard.edu>
Page 13:
The 2022 CES survey was conducted over the Internet by YouGov. The Common Content was asked of 60,000 adults interviewed in September–November 2022 (for pre-election data), and in November–December 2022 (for post-election data). The sampling method uses YouGov’s matched random sample methodology.
[7] Textbook: Mind on Statistics (4th edition). By Jessica M. Utts and Robert F. Heckard. Brooks/Cole Cengage Learning, 2012.
Pages 164–165:
Surveys that simply use those who respond voluntarily are sure to be biased in favor of those with strong opinions or with time on their hands. …
According to a poll taken among scientists and reported in the prestigious journal Science … scientists don’t have much faith in either the public or the media. … It isn’t until the end of the article that we learn who responded: “The study reported a 34% response rate among scientists, and the typical respondent was a white, male physical scientist over the age of 50 doing basic research.” … With only about a third of those contacted responding, it is inappropriate to generalize these findings and conclude that most scientists have so little faith in the public and the media.
[8] “Guide to the 2022 Cooperative Election Study.” By Stephen Ansolabehere, Brian Schaffner, and Marissa Shih. Harvard/YouGov Cooperative Election Study, August 2023. <dataverse.harvard.edu>
Page 13:
Sampling and Sample Matching
Sample matching is a methodology for selection of “representative” samples from non-randomly selected pools of respondents. It is ideally suited for Web access panels, but could also be used for other types of surveys, such as phone surveys. Sample matching starts with an enumeration of the target population. For general population studies, the target population is all adults, and can be enumerated through the use of the decennial Census or a high quality survey, such as the American Community Survey. …
The purpose of matching is to find an available respondent who is as similar as possible to the selected member of the target sample. The result is a sample of respondents who have the same measured characteristics as the target sample. Under certain conditions, described below, the matched sample will have similar properties to a true random sample. That is, the matched sample mimics the characteristics of the target sample. It is, as far as we can tell, representative of the target population (because it is similar to the target sample).
Page 15:
Weighting
The sample is weighted to adjust for any remaining imbalance that exists among the matched sample. Such imbalance results from the fact that the closest match for a particular individual from the target sample is not necessarily a perfect match across all demographics. The matched cases and the frame were combined and the combined cases were balanced on multiple moment conditions using the politically representative citizen frame.
[9] Paper: “Do Non-Citizens Vote in U.S. Elections?” By Jesse T. Richman, Gulshan A. Chattha, and David C. Earnest. Electoral Studies, December 2014. Pages 149–157. <www.sciencedirect.com>
Page 151:
It is impossible to tell for certain whether the non-citizens who responded to the survey were representative of the broader population of non-citizens, but some clues can be gained by examining education levels. … We confront this issue primarily by weighting the data.
Throughout the analysis (with the exception of the appendix) we report results produced from weighted data. Weight construction began with CCES [Cooperative Congressional Election Study] case weights, but then adjusted these by race to match the racial demographic of the non-citizen population. Our concern with using regular CPS [Current Population Survey] case-weights was that weights were constructed based upon overall demographic characteristics without attention to the demographic character of the non-citizen population. … Weighting produces a non-citizen sample that appears to be a better match with Census estimates of the population. For instance, 32.5 percent of the weighted sample had no high school degree. …
… Among the 337 immigrant non-citizens who responded to the CCES, 50 (14.8%) indicated in the survey that they were registered.
[10] Email from Dr. Jesse Richman to Just Facts, April 16, 2024:
“Voter Registration Status … Yes [=] 32 … No [=] 441 … Don’t know [=] 18”
CALCULATION: 32 / (32 + 441) = 6.8%
[11] Poll: “National Hispanic Survey Results.” By John McLaughlin. McLaughlin & Associates, June 21, 2013. <mclaughlinonline.com>
Page 3:
This bi-lingual national survey of 800 Hispanics was conducted from June 5th through June 16th, 2013.
Interview selection was within predetermined census units of Hispanic adults. 560 interviews were conducted via landline telephone by professional interviewers. To increase coverage, this landline sample was supplemented with 240 interviews, 30%, conducted via internet of cellphone only users. 64% of all respondents use cell phones. 60% of all interviews were conducted in Spanish. 93% of all respondents speak at least some Spanish at home. These samples were then combined and structured to correlate with actual adult Hispanic census population.
Page 4:
The uniqueness of this poll is that it is very strong demographically and methodologically. 60% of the interviews were actually conducted in Spanish; 76% speak Spanish mostly or equally. 23% always speak Spanish; 93% speak at least some Spanish at home; 30% of the interviews were conducted among cell phone only users. 64% of Hispanic adults have cell phones.
Page 68: “Voter Profile … Non-Citizen … Registered [=] 13%”
NOTE: Credit for bringing this poll to attention belongs to Rowan Scarborough of the Washington Times.
[12] Email from Dr. Jesse Richman to Just Facts, April 16, 2024:
“Voter Registration Status … Yes [=] 32 … No [=] 441 … Don’t know [=] 18”
CALCULATION: 32 / (32 + 441) = 6.8%
[13] “Expert Report of Jesse T. Richman in the Case of Mi Familia Vota v. Fontes.” October 13, 2023. <www.justfacts.com>
Page 72:
Each year the CES [Cooperative Election Study (CES) survey (formerly known as the Cooperative Congressional Election Study or CCES)] asks respondents their citizenship status with at least one question (and more recently with two questions) and asks respondents whether they were registered to vote or not. In addition, the CES long retained the voter file firm Catalist to match survey respondents with voter-file and commercial records in an effort to verify respondents’ statements of whether they were registered to vote. In 2022 the CES switched its voter file matching from Catalist to instead use the TargetSmart database.
[14] Calculated with data from the paper: “Do Non-Citizens Vote in U.S. Elections?” By Jesse T. Richman, Gulshan A. Chattha, and David C. Earnest. Electoral Studies, December 2014. Pages 149–157. <www.sciencedirect.com>
Page 150: “Of 339 non-citizens identified in the 2008 survey, Catalist matched 140 to a commercial (e.g. credit card) and/or voter database.”
CALCULATION: 140 / 339 = 41%
[15] Email from Just Facts to Dr. Jesse Richman, April 16, 2024:
What portion of the people who identified themselves as non-citizens in the 2022 CES were matched in the TargetSmart database?
Email from Dr. Jesse Richman to Just Facts, April 16, 2024:
It was 7.21 percent. … Unlike Catalist (especially in the earlier years of the CCES) targetsmart doesn’t seem to be matching on anything other than the voter registration list. Thus, if someone is matched by them, it’s mostly because they are registered to vote.
[16] Report: “Beyond the Core and Periphery: A New Look at Voter Participation Across Elections.” By Stephen Ansolabehere and Brian Schaffner. Harvard/YouGov Cooperative Congressional Election Study, November 30, 2015. <cces.gov.harvard.edu>
Pages 3–4:
The second source of data we analyze in this paper is a 1% sample (N=2,969,951) of American adults from Catalist. Catalist is a prominent voter file firm typically providing data to progressive clients, but also to academic institutions under a special subscription plan. Catalist maintains a database of information for nearly every American adult. The database is first built on voter file records acquired from all 50 states, and then a variety of different information is appended to that data from marketing firms, census data, and other sources. Catalist also builds several of its own models into the dataset to distinguish, for example, the partisanship, ideology, and political activism of each individual. The Catalist dataset is ideal for studying voter turnout because it includes precise information on each individual’s turnout behavior and is sufficiently large so that even relatively uncommon voter types can still be studied with precision.
[17] Webpage: “Catalist Data.” Catalist. Accessed November 15, 2016 at <bit.ly>
Catalist provides data and data-related services to progressive organizations to help them better identify, understand, and communicate with the people they need to enhance, persuade, and mobilize.
COMPREHENSIVE
Our national database contains more than 240 million unique voting-age individuals. We start with a rich, reliable foundation of 185 million registered voters—a unified national voter file—collected from Election Officials in all 50 states and the District of Columbia.
Commercial data adds another 55 million unregistered individuals and enhances the entire database with hundreds of fields, including household attributes, purchasing and investment profiles, donation behavior, occupational information, recreational interests, and engagement with civic and community groups.
We also integrate data from the Census, specialty data, and media market geographies. We have records going back more than a decade, giving our clients the unique ability to see both short-term and long-term trends.
ACCURATE
Catalist data is continually updated and audited by data quality experts. Catalist’s Data Operations team adheres to a rigorous multi-step quality assurance process that ensures that each update improves the quality of the database.
THE CATALIST NATIONAL DATABASE BY THE NUMBERS:
240 Million+ unique voting-age individuals
185 Million+ registered voters
55 Million+ unregistered voting age persons
NOTE: 240 million voting-age individuals = 96.3% of the U.S. adult population of 249,291,898 people in 2016 (when this Catalist webpage was accessed). [Calculated with the dataset: “Annual Estimates of the Resident Population by Single Year of Age and Sex for the United States: April 1, 2010 to July 1, 2019.” U.S. Census Bureau. Accessed May 9, 2024 at <www.census.gov>]