The USC Center for the Study of Immigrant Integration (CSII) estimates are based on a pooled sample of the 2010-2014 American Community Survey (ACS) microdata from IPUMS-USA, University of Minnesota, www.ipums.org. The methodology involves first using the ACS data to estimate the total undocumented population in the U.S.; for detailed methodology on how we do that, including the use of conditional screens, country controls, and a stratified probability model, see here. To determine who was DACA-eligible from the pool of undocumented residents, we first applied the age, age at arrival, educational requirements, and continuous residency criteria given in the 2012 executive order, and developed an estimate of who was initially eligible in 2012. We then followed a strategy similar to the Migration Policy Institute and “aged in” an additional four years of people who met the initial screen, applying an estimated drop-out rate to exclude those who might now meet the age criteria but not the educational requirements.
Once those eligible for DACA were determined, the number of actual recipients was estimated using a combination of an additional conditional screen and an Iterative Proportional Fitting (IPF) process. The screen involved education: we assumed that all those DACA eligible who had completed or were in grad school have applied for DACA and we assumed that the vast majority of those in college and nearly all of those with a B.A. had applied for and received DACA. The resulting number is well below the implicit numbers estimated by UC San Diego scholar Tom Wong in his most recent DACA survey; because there is reason to believe that the sampling strategy in that approach may over-sample the educated and also reason to believe that our “aging in” approach misses current college experience, this is a reasonable outcome. We then used an IPF process on the rest of DACA eligibles to match published data from the U.S. Citizen and Immigration Services on the total number of DACA approvals by state and country of origin as of the end of the third quarter of fiscal year 2017 (through June 30th 2017), available here. We chose to make the final fit tighter on country than on state because recipients can change their state of residence but they cannot change their country of origin.
The last step of methodology involved assigning the individual-level estimates of the DACA eligible and DACA recipients to Congressional Districts (115th Congress). To do so, we used a 2010 population-based crosswalk between Public Use Microdata Areas (PUMAs) and Congressional Districts from Mable/GeoCorr12. A geographic crosswalk was necessary because the ACS microdata includes only very limited geographic information, with the PUMA being lowest level of geography attached to individuals. PUMAs contain a population of at least at least 100,000 while Congressional Districts have an average population of about 700,000. We randomly assigned individuals to Congressional Districts in proportion to the 2010 population distribution, and then aggregated the results by Congressional District to derive the final estimates presented in the interactive map. While this introduces some unknown degree of geographic error in the resulting estimates, it is likely to be relatively minimal given that most PUMAs are either entirely contained or largely contained in a single Congressional District. To caution users of the map against utilizing unreliable estimates, we include an asterisk and a note alongside the data for Congressional Districts with fewer than 50 unweighted survey respondents identified as DACA eligible. We also do not report any data for Congressional Districts in which our estimates showed fewer than 250 DACA eligible.
Estimated GDP loss from removing DACA recipients from the workforce by Congressional District is from Center for American Progress analysis of the underlying CSII estimates. State-level estimates of GDP loss per DACA worker were applied to the estimated number of DACA workers by congressional district (91 percent of the DACA recipients in each Congressional District were assumed to be workers). For more on the methodology, see Nicole Prchal Svajlenka, Tom Jawetz, and Angie M. Bautista-Chavez, "A New Threat to DACA Could Cost States Billions of Dollars" (Washington: Center for American Progress, 2017), available at https://www.americanprogress.org/issues/immigration/news/2017/07/21/436419/new-threat-daca-cost-states-billions-dollars/. GDP loss is annual in 2013 dollars and is only reported for Congressional Districts in states with at least 500 estimated DACA recipients. Estimated GDP losses should be viewed as conservative for three reasons. The first is simply that they are based on GDP in 2013, and are reported in 2013 dollars. The second is that they assume the skill distribution of DACA workers reflects the broader unauthorized workforce, yet we know that DACA recipients are likely, given the requirements, to be more educated than the norm for unauthorized workers. The third is that many working DACA recipients are also in school and are likely to increase their income and productivity in the future (as are those who are not working, but in school).