Law enforcement detains a protester near an Immigration and Customs Enforcement facility in Broadview, Ill., Friday, Oct. 3, 2025. (AP Photo/Erin Hooley)
From the outset of the second Trump administration, it has become apparent that the administration is
operating by a different playbook than its first time around. But how different has it actually been?
I conducted an exploratory analysis of individual-level administrative arrest and detention data from
the U.S. Immigration, Customs, and Enforcement Agency (ICE), as well as recent data obtained through a
Freedom of Information Act request by the Deportation Data Project. This project aims to evaluate
changes between Trump's first and second
administrations across four dimensions: volume of arrests, geographic targeting, demographic
composition, and criminal backgrounds of those arrested.
I used FY2017 and FY2018 ICE arrests data, and 2025 ICE arrests and detentions data obtained through
the FOIA request. The 2025 data include administrative arrests from September 2023 through
mid-October 2025. The raw data files are too large to be kept in the GitHub repository, but can be
downloaded from the website.
I used Python (pandas, numpy, BeautifulSoup) through Colab and VSCode and Vega-Lite to clean data
and
create the charts in this project. All my workbooks and .json files can be found in the repository.
I cleaned for duplicate entries by removing rows that had identical unique identifiers and dates. I
also
removed unknown/missing data for a number of the charts – notes for replication can be found under
each section.
The chart below maps seven-day rolling averages of arrests for the first and second Trump administrations before and after inauguration. In both administrations, daily ICE arrests immediately increased post-inauguration, but the change is much more dramatic in the second administration. Moreover, whereas daily arrests stayed relatively stable during the first months of Trump 1.0, daily arrests seem to trend upwards in Trump 2.0. Spikes in the 2025 data correspond with major ICE operations conducted during those times, namely in the Los Angeles, Washington D.C., and Chicago areas.
The Criminality field in the 2017 arrest data was created with the “Most Serious Criminal Charge Status” field, recoding “Turned Over to INS without Prosecution,” “Overturned,” “Dismissed,” and missing values as “No Criminal Charges.”
The top nationality of people arrested by ICE continues to be Mexico, but its share has decreased by 25%. Absent from the top 15 in 2017, Venezuela now ranks in the top five. All four nationalities that had access to temporary protection and work authorization under the CHNV (Cuba, Haiti, Nicaragua, Venezuela) program, which Trump abruptly ended in 2025, are represented in the top 15.
The shapefiles for the Area of Responsibility (AOR) map are from the Deportation Data Project. I used mapshaper to simplify and convert them to TopoJSONs.
Although the distribution of arrests across age groups is similar between the two administrations, the Trump Administration has arrested nearly three times the number of children. The youngest child arrested in the first months of Trump 2.0 was less than a year old.
I removed entries that had “Unknown” listed for gender (~6000) and missing values for birth year (1).
In the past, ICE typically has relied on criminal record data and partnerships with local law enforcement to collect information like immigration status and physical addresses to make arrests. However, during major operations, ICE is more likely to use indiscriminate methods of arrests - such as targeting workplaces and public areas. For example, in the beginning of September, the share of arrests of people with no criminal record in the Washington D.C. area rose to over 70%.
Fields on crime are missing from 2025 arrest data, while AOR is absent from 2025 detention
data. I joined these files using shared unique identifiers. The 2017 datasets lacked a
shared identifier but arrest data included crime fields. Some Trump 1.0 arrests shown may
not have resulted in detentions.
I scaraped the National Crime Information Center Code Manual from here.
The analysis reveals that Trump 2.0 differs from the first administration on a number of dimensions:
the sheer number of daily arrests is up, the nationalities that make up the arrests are different,
and new U.S. regions are being targeted. Furthermore, the share of people without criminal records
that ICE has picked up during its raids has increased dramatically.
There are limitations to the conclusions that this project can make about the administration’s
targeting of nationalities and regions, since the demographics of the immigrant population have
changed between administrations. Further analyses should take these demographic changes into
consideration.
Despite these limitations, it is still abundantly clear from the analysis that the administration is
a far cry from its claim of only arresting the "worst of the worst." In fact, even
though it has
detained nearly twice as many people in its first months, it has arrested fewer people who have
committed violent crimes compared to its first term. The project provides a replicable framework to
monitor the administration as additional data become available and as the administration's policies
continue to evolve.