Data Analysis Objective
Recent events in our world have left many people scared and confused. It's a difficult time for all of us. My go to for coping with a difficult situation is to perform data analytics so I can be confident in my stances and contribute however I can to the discussion. This is what I aimed to do a week ago when I downloaded data directly from https://www.ice.gov/detain/detention-management for Fiscal Year 2019 - Jan 8th 2026. I then spent the last week or so going through the tedious process of reconciling varying data structures over the years, attempting to understand the data point definitions, and consolidating the data for ease of public consumption.
Initial Questions and Data Prep
I started my exploration of the data with several questions.
- How many people are being detained, according to the official government data?
- What portion of these people are actually criminals?
- Who is being released, and at what rate?
To start this process I needed to identify what data sets available in the downloads would help me answer some of these questions. I found the following data sets to be data rich and applicable to my questions.
- ICE Average Length of Stay by Arresting Agency, Month and Criminality
- ICE Final Releases by Release Reason and Criminality
- ICE FACILITIES DATA
I started by reviewing each Fiscal Year's data (A fiscal year running from October - September) and comparing the column headers between years to know how I could combine the data for over time analysis. There were notable differences in data structure between FY19-20, FY21-24, and FY25-26. This was unsurprising due to changes of administration during the date ranges. I want to emphasize that although the data structures changed, it appeared most data was still preserved but just categorized differently. Either way this still required manual review of columns and creation of reconciled column headers that I would use in the combining of all data.
Average Daily Persons by Fiscal Year and Criminality
The first data points I examined over time were Average Daily Persons (also known as ADP) and Criminality. This was included in the by facility data with more granularity than the general ADP data. This value represents the average amount of detainees per day in facilities. In an effort to provide a richer, more data compact visualization, I choose to aggregate this data by the Criminal or Non-Criminal flag in the data set. The Criminal flag indicates detainees that are convicted criminals, regardless of the severity of the charge. The Non-Crim flag indicates individuals detained for civil immigration reasons.
Important Note: The FY2026 Data is incomplete, as this data was last updated on Jan 8th 2026. The ADP values represent the sum of ADP values across facilities to give us an idea of how many individuals are detained in a facility on average per day per fiscal year.
Average Daily Persons by Fiscal Year and ICE defined Threat Levels
The next pair of data points I compared were the ADP and the ICE defined threat levels. Similar to the Criminality this is represented over fiscal years.
There is a notable trend, represented by both ICE defined threat levels and Crim/non-Crim categorizations that indicates the vast majority of detainees are civil immigration issues as opposed to violent convicted criminals.
Total Detention Estimates
Beyond the average daily persons and their applicable criminality I wanted to know how many people actually pass through ICE detainment. This was a much more challenging data point to isolate as the data set from ICE does not explicitly call out throughput of detainees. I did find that in addition to ADP, the data set included a Average Length of Stay (ALOS) data point. This data point is entered for individuals on their release which means any attempts to calculate the actual amount of detainees over a year using ALOS and ADP is, at best, an estimate. In addition, if there are many who are detained for extended amounts of time and are not released then the data point would remain un-reported for that individual, which means the estimates are likely to be low estimates.
Regardless, I was able to extract estimates by taking the (ADP x 365) / ALOS which gives an estimate of the total population that was detained and released over a fiscal year.
Since the data set I downloaded from the official government database did not include a full 2026 fiscal year data set, and the file was named FY26_detentionStats01082026, its reasonable to assume the last time it was updated was January 8th 2026. Using this partial data set I was able to produce a forecast for FY2026 which assumes consistent ALOS and ADP throughout the year.
This estimate produces an estimate of approximately 1.1 million individuals with 78% of them being classified as non-crim, by the end of FY2026.
Release
The sizable increase to approximately 1.1 million individuals in FY2026 compared to previous years leaves another lingering question. Where are these individuals going after detainment, how many are being deported, how many are being dismissed. The data set did not provide clear indicators of which release reason meant which however after reviewing their footnotes, and consulting with a GenAI, the most likely deportation category is Release to Remove. I was able to produce a chart, over fiscal years, of the varying release reasons. In addition, by using the release to date for the data set in FY2026 (about 100 days into the FY) I multiplied the values by 3.65 to get an estimate of the total fiscal year results, again this assumes continued rates matching the FY2026 rates so far.
Conclusion, for now.
These estimates provide several stark indicators about the increase in ICE detainment, the proportion of Criminally charged detainees vs civil immigration issues, and the rate of deportations. I hope that my analytics have helped provide some clarity of the state of our country in these challenging times. I intend to continue exploring the data set, searching for additional data sets, and periodically checking for newly released data in the future. For now, I felt that it was important to get what I had already out there.
Comments