Unreported Detention: Un- and under-reported ICE facilities

An analysis of ICE detention statistics not included in the FY 2025 annual detention report

Comparing the Deportation Data Project’s daily detention counts with ICE’s published annual statistics to identify unreported facilities and track population changes

Author

Carwil Bjork-James

Published

April 10, 2026

Overview

This page offers an analysis of two forms of data on the population of people held by US Immigration and Customs Enforcement (ICE): summary data on detention published by ICE on its website and a detailed and granular dataset of arrests, removals, and overnight populations under ICE custody, released under the Freedom of Information Act and archived by the Deportation Data Project.

The former data, is published online the ICE Detention Management Division. I have also produced a map of reported facilities based on sixteen years of reporting (eight annual reports and two summary sheets) from ICE.

The newer data covers “every ICE encounter, detainer request, arrest, book-in to detention, and removal” between September 1, 2022 and October 15, 2025.

As becomes clear below, the more extensive data shows that ICE’s public reporting is incomplete, and the uncounted population grew massively during 2025, from hundreds of individuals per night in the final months of the Biden administration to as many as forty-five hundred individuals, and frequently in excess of three thousand in summer and early fall 2025. This unreported network includes over ICE-owned holding facilities, nine dedicated ICE detention centers, local jail and state prisons, federal prisons, and over 200 medical facilities.

The FOIA-supplied data also exceeds ICE’s annual reporting in those facilities that do match by 1,038 detainees per night over the course of FY 2025, though this only amounts to 2% of the total population.

In much oyf this report, facility-level summaries distinguish between facilities with Peak Population ≥ 2 and those with Peak Population < 2. Facilities with no more than one detainees at a time account for many DDP facility codes but negligible population. Separating them keeps attention on facilities with substantive detention activity.

This page is part of a larger data project to gather, summarize, and analyze public data about immigration enforcement in the United States. The focus here is detention facilities used by US Immigration and Customs Enforcement (ICE).

Code

library(dplyr)
library(tidyr)
library(ggplot2)
library(scales)
library(knitr)
library(sf)
library(leaflet)
library(leafpop)
library(kableExtra)
library(leaflet.extras)

Code

# Pre-computed data from the ice-detention pipeline.
# Run copy-data.sh to refresh from the pipeline project.
ice_fy25          <- readRDS("data/ice_fy25.rds")
ddp_fy25          <- readRDS("data/ddp_fy25.rds")
ddp_fy25_keyed    <- readRDS("data/ddp_fy25_keyed.rds")
daily_totals      <- readRDS("data/daily_totals.rds")
unmatched         <- readRDS("data/unmatched.rds")
daily_unmatched   <- readRDS("data/daily_unmatched.rds")
monthly_unmatched <- readRDS("data/monthly_unmatched.rds")
peak_fy25         <- readRDS("data/peak_fy25.rds")
biden_unmatched   <- readRDS("data/biden_unmatched.rds")
trump_unmatched   <- readRDS("data/trump_unmatched.rds")
daily_unreported  <- readRDS("data/daily_unreported.rds")

Aggregate comparison

The DDP data (Deportation Data Project) provides daily facility-level detention counts. To compare with the ICE FY25 annual statistics, we average DDP daily counts from October 1, 2024 through September 24, 2025—the date stamp on the FY25 spreadsheet. On average, the DDP data reflects 2581 more detainees per day than reported in the ICE FY25 annual statistics. As we’ll see below, about 1,000 of these detainees are in facilities that appear in both sources but with higher counts in the DDP data. The rest are in facilities that do not appear at all in the ICE FY25 spreadsheet, accounting for a full-year average of roughly fifteen-hundred detainees per day. However, since the unmatched population grew dramatically during 2025, the daily unmatched population was much higher towards the end of the year.

Code

tibble(
  Source      = c("ICE FY25 annual stats", "DDP daily average"),
  Facilities  = c(nrow(ice_fy25), nrow(ddp_fy25)),
  `Total ADP` = c(round(sum(ice_fy25$ice_adp, na.rm = TRUE)),
                   round(sum(ddp_fy25$ddp_adp)))
) |>
  kable()

Source	Facilities	Total ADP
ICE FY25 annual stats	189	47549
DDP daily average	768	50130

The DDP covers 768 facility codes while the ICE spreadsheet lists 189 facilities. The DDP’s higher total reflects the many additional facilities (hold rooms, federal prisons, medical facilities) not included in the ICE annual statistics.

Daily population over time

Code

fy_breaks <- as.Date(c("2023-10-01", "2024-10-01", "2025-10-01"))
fy_labels <- data.frame(
  x = as.Date(c("2024-04-01", "2025-04-01", "2025-10-08")),
  label = c("FY24", "FY25", "FY26")
)

ggplot(daily_totals, aes(x = date, y = total_pop)) +
  geom_line(linewidth = 0.3) +
  geom_vline(xintercept = fy_breaks, linetype = "dashed", alpha = 0.4) +
  # Annual averages as calculated as follows
  # facilities_aggregated$FY24$sum_criminality_levels |> sum()
  # 36276.8
  # facilities_aggregated$FY25$sum_criminality_levels |> sum()
  # 47550.2
  # facilities_aggregated$FY26$sum_criminality_levels |> sum() # taken with February 2026 data
  # 66811.9
  # geom_segment(x=as.Date("2023-10-01"), xend=as.Date("2024-10-01"), 
  #              y=36276.8, linetype = "dashed", alpha = 0.4, linewidth=0.1, color = "blue") +
  # geom_segment(x=as.Date("2024-10-01"), xend=as.Date("2025-10-01"),
  #              y=47550.2, linetype = "dashed", alpha = 0.4, linewidth=0.1, color = "blue") +
  # geom_segment(x=as.Date("2025-10-01"), xend=as.Date("2025-11-01"),
  #              y=66811.9, linetype = "dashed", alpha = 0.4, linewidth=0.3, color = "blue") +
  geom_text(data = fy_labels, aes(x = x, y = max(daily_totals$total_pop) * 1.04,
                                   label = label), inherit.aes = FALSE, size = 3) +
  scale_y_continuous(labels = comma) +
  labs(x = NULL, y = "Total detained population")

Figure 1: Total ICE detained population per day (DDP data, all facilities)

Facility-level matching

ICE uses detainee locator codes, or DETLOCs to keep track of which facilities detainees are housed in. These codes are included in 2015 and 2017 facility lists (describing ICE detention populations from FY 2010 to 2017), in a list maintained by the Deportation Data Project, and in separate lists compiled by the Vera Institute and The Marshall Project. Our data wrangling matched the FY 2019-2026 data to these other lists by exact facility name, address, city and state, and other criteria. All 189 detention facilities listed in the ICE spreadsheet for FY2025 have one or more known DETLOC.

We join the two datasets through canonical facility IDs. Each DDP facility code is mapped to a canonical ID via detloc_lookup_full, which preserves all DETLOC variants from every source (DMCP, DDP, hold/ERO, Vera). Some canonical facilities have multiple codes—for example, Central Arizona Florence Correctional Complex uses CCADCAZ (DMCP) and CCAFLAZ (DDP), and Delaney Hall uses ESSEXNJ (DMCP) and DHDFNJ (DDP). Using the full lookup ensures all codes contribute ADP to the correct facility.

Code

# Aggregate DDP by canonical_id for matched facilities.
ddp_by_canonical <- ddp_fy25_keyed |>
  filter(in_ice_fy25) |>
  group_by(canonical_id) |>
  summarise(ddp_adp = sum(ddp_adp), .groups = "drop")

comparison <- ice_fy25 |>
  # ICE FY25 can have multiple rows per canonical_id (e.g., Adelanto + Desert View Annex)
  group_by(canonical_id) |>
  summarise(
    canonical_name = first(canonical_name),
    ice_adp = sum(ice_adp, na.rm = TRUE),
    .groups = "drop"
  ) |>
  full_join(ddp_by_canonical, by = "canonical_id") |>
  mutate(
    in_ice = !is.na(ice_adp),
    in_ddp = !is.na(ddp_adp),
    diff   = ddp_adp - ice_adp,
    pct_diff = diff / ice_adp * 100
  )

matched <- comparison |> filter(in_ice & in_ddp)

tibble(
  Category = c("In both sources", "ICE only (no DDP match)",
               "Matched: ICE total ADP", "Matched: DDP total ADP",
               "Matched: aggregate difference"),
  Value = c(
    sum(comparison$in_ice & comparison$in_ddp),
    sum(comparison$in_ice & !comparison$in_ddp),
    round(sum(matched$ice_adp, na.rm = TRUE)),
    round(sum(matched$ddp_adp, na.rm = TRUE)),
    paste0(round(sum(matched$ddp_adp, na.rm = TRUE) -
                   sum(matched$ice_adp, na.rm = TRUE)),
           " (", round((sum(matched$ddp_adp, na.rm = TRUE) /
                           sum(matched$ice_adp, na.rm = TRUE) - 1) * 100, 2), "%)")
  )
) |>
  kable()

Category	Value
In both sources	188
ICE only (no DDP match)	0
Matched: ICE total ADP	47549
Matched: DDP total ADP	48586
Matched: aggregate difference	1037 (2.18%)

Facility-level agreement

Code

both_nonzero <- matched |> filter(ice_adp > 0, ddp_adp > 0)

ggplot(both_nonzero, aes(x = ice_adp, y = ddp_adp)) +
  geom_abline(slope = 1, intercept = 0, linetype = "dashed", alpha = 0.5) +
  geom_point(alpha = 0.6) +
  scale_x_log10(labels = comma) +
  scale_y_log10(labels = comma) +
  labs(x = "ICE FY25 ADP (annual stats)", y = "DDP mean daily population")

Figure 2: FY25 ADP comparison by facility (log scale). Dashed line = perfect agreement.

Code

tibble(
  `Agreement band` = c("Within 10%", "Within 25%", "Within 50%", ">50% difference", "Annual report shows greater value"),
  Facilities = c(
    sum(abs(both_nonzero$pct_diff) <= 10),
    sum(abs(both_nonzero$pct_diff) > 10 & abs(both_nonzero$pct_diff) <= 25),
    sum(abs(both_nonzero$pct_diff) > 25 & abs(both_nonzero$pct_diff) <= 50),
    sum(abs(both_nonzero$pct_diff) > 50),
    sum(both_nonzero$pct_diff < 0)
  )
) |>
  kable()

Agreement band	Facilities
Within 10%	115
Within 25%	26
Within 50%	30
>50% difference	17
Annual report shows greater value	24

Among the 188 matched facilities with nonzero ADP in both sources, 115 agree within 10%. The log-scale correlation is 0.998. There are 24 facilities with a lower count in the DDP data and the largest percentage difference in this direction is -4.2%.

Table 1: ICE FY25 facilities with no DDP match

Code

ice_only <- comparison |> filter(in_ice & !in_ddp)
if (nrow(ice_only) > 0) {
  ice_only |>
    select(canonical_id, canonical_name, ice_adp) |>
    mutate(ice_adp = round(ice_adp)) |>
    arrange(desc(ice_adp)) |>
    kable(col.names = c("ID", "Facility", "ICE ADP"))
}

All ICE FY25 facilities have a matching DDP facility code.

Unmatched facilities

576 of 768 DDP facility codes do not appear in the ICE FY25 annual statistics. These are facilities where ICE holds detainees but which are excluded from the published spreadsheet.

By facility type

Facility types are drawn from the Vera Institute’s classification system. “Dedicated” facilities are purpose-built or contractually dedicated to ICE detention. “Non-Dedicated” are county jails and other facilities that also hold ICE detainees. “Hold/Staging” are short-term ERO processing rooms. “Federal” includes BOP prisons and USMS facilities.

Code

unmatched |>
  group_by(type_grouped, peak_class) |>
  summarise(
    Facilities = n(),
    `FY25 ADP` = round(sum(ddp_adp)),
    .groups = "drop"
  ) |>
  arrange(desc(`FY25 ADP`)) |>
  kable(col.names = c("Facility type", "Peak class", "Facilities", "FY25 ADP"))

Table 2: DDP facilities not in ICE FY25, by Vera facility type and ADP class

Facility type	Peak class	Facilities	FY25 ADP
Hold/Staging	Peak ≥ 2	169	1163
Dedicated	Peak ≥ 2	2	170
Unclassified	Peak ≥ 2	135	140
Federal	Peak ≥ 2	24	31
Non-Dedicated	Peak ≥ 2	23	30
Unclassified	Peak < 2	185	6
Non-Dedicated	Peak < 2	14	2
Medical	Peak ≥ 2	1	1
Family/Youth	Peak < 2	1	0
Family/Youth	Peak ≥ 2	1	0
Federal	Peak < 2	13	0
Hold/Staging	Peak < 2	7	0
Medical	Peak < 2	1	0

The 576 unmatched facilities account for a full-year ADP of roughly 1,544. There were 475 facilites that averaged fewer than 2 detainees per day, contributing 136 ADP in total. Among them, 221 facilities never had more than 1 detainee per day, contributing 8 ADP in total.

Daily and monthly detention in unmatched facilities

Code

ggplot(daily_unmatched, aes(x = date, y = total_pop)) +
  geom_line(linewidth = 0.3) +
  scale_y_continuous(labels = comma) +
  scale_x_date(date_labels = "%b %Y", date_breaks = "2 months") +
  labs(x = NULL, y = "Detained population (unmatched facilities)")

Figure 3: Daily detained population in facilities not in ICE FY25 annual statistics

Code

monthly_by_type <- monthly_unmatched |>
  group_by(month, type_grouped) |>
  summarise(daily_avg = sum(n_detained) / n_distinct(date), .groups = "drop")

ggplot(monthly_by_type, aes(x = month, y = daily_avg, fill = type_grouped)) +
  geom_col() +
  scale_y_continuous(labels = comma) +
  scale_x_date(date_labels = "%b %Y", date_breaks = "2 months") +
  labs(x = NULL, y = "Average daily population", fill = "Facility type")

Figure 4: Average daily population in unmatched facilities, by month and type

Last 60 days of FY25

The final 60 days of the FY25 data window (July 27 – September 24, 2025) capture the period of peak detention.

Code

last60_start <- as.Date("2025-07-27")
last60_end   <- as.Date("2025-09-24")

last60 <- monthly_unmatched |>
  filter(date >= last60_start, date <= last60_end)

non_empty <- last60 |>
  group_by(detention_facility_code) |>
  summarise(total_detained = sum(n_detained), .groups = "drop") |>
  filter(total_detained > 0)

last60_by_type <- last60 |>
  filter(detention_facility_code %in% non_empty$detention_facility_code) |>
  group_by(type_grouped) |>
  summarise(
    Facilities = n_distinct(detention_facility_code),
    `Daily avg` = round(sum(n_detained) / n_distinct(date)),
    .groups = "drop"
  ) |>
  arrange(desc(`Daily avg`))

last60_total <- sum(last60_by_type$`Daily avg`)

last60_by_type |>
  rename(`Facility type` = type_grouped) |>
  kable()

Table 3: Average daily population in unmatched facilities, last 60 days of FY25

Facility type	Facilities	Daily avg
Hold/Staging	162	1987
Dedicated	2	692
Unclassified	180	334
Federal	30	75
Non-Dedicated	22	52
Medical	2	2
Family/Youth	1	0

In the last 60 days, 3,142 people per day were held in facilities absent from the ICE FY25 annual statistics.

Peak FY25 population by facility

This table of the facilities with the largest peak population in FY 2025 help us make sense of th unreported population. A variety of different kinds of facilities play a roll:

New Federal facilities like the El Paso Soft-Sided Facility (ELPASOFS), which is distinct from the El Paso Processing Center (EPC) and Camp East Montana (EROFCB), held large numbers of detainees. This new facility was not added to the FY2025 statistics and was supplanted by Camp East Montana before the new fiscal year. The Ursula Centralized Processing Center, formally associated with Customs and Border Patrol, is listed as holding an average of 32 and a mamximum of 463 ICE detainees.
Dedicated ICE detention centers supplied by states, like the Baker Correctional Institution (FLBAKCI), a Florida state prison that was turned over to use by ICE in 2025, and held as many as 294 prisoners in FY 2025. The Florida Soft-Sided Facility (FLDSSFS, also known as Alligator Alcatraz was the sceond largest unreported facility, with a peak population of 1,482 detainees on July 30, 2025. These facilities were added to the FY 2026 statistics but left off the 2025 document.
Holding facilities associated with airports like the Krome Hold Room (KROHOLD) near Miami International Airport. This was one of the few facilities used at a large scale in the final days of the Biden administration. The Arizona Removal Operations Coordination Center (AROCC) operates at the Phoenix-Mesa Gateway Airport. As with other hold rooms, these facilities were reportedly exceeding the 12-hour limit on detention time.
Field Offices and Sub-Offices of the Enforcement and Removal Operations (ERO) division of ICE, which are not designed for long-term detention but which held large numbers of detainees during the surge in 2025. This shift coincided with new ICE guidance authorizing detention up to 72 hours, but preceded the new permission and exceeded its limits.
A handful of local prisons were used to hold ICE detainees en masse, including the Richwood Correctional Center (RICHWOOD) in Louisiana, which held up to 1,200 detainees at a time.

Code

peak_fy25 |>
  arrange(desc(peak_pop)) |>
  head(20) |>
  select(detention_facility, state, type_grouped, peak_pop, peak_date, mean_pop) |>
  kable(col.names = c("Facility", "State", "Type", "Peak", "Peak date", "Mean ADP"))

Table 4: Top 20 unmatched facilities by FY25 peak detained population

Facility	State	Type	Peak	Peak date	Mean ADP
EL PASO SOFT SIDED FACILITY	TX	Hold/Staging	1606	2025-08-12	243.2
FLORIDA SOFT-SIDED FACILITY-SOUTH	FL	Dedicated	1482	2025-07-30	170.3
KROME HOLD ROOM	FL	Hold/Staging	600	2025-02-17	36.5
URSULA CENTRALIZED PROCESSING CNTR	TX	Unclassified	463	2025-07-11	31.6
LOS CUST CASE	CA	Hold/Staging	455	2025-07-11	58.7
BAKER C. I.	FL	Unclassified	294	2025-09-24	15.7
RICHMOND SUB-OFFICE HOLD	VA	Hold/Staging	277	2025-09-07	16.2
BROADVIEW SERVICE STAGING	IL	Hold/Staging	245	2025-09-20	31.7
NYC HOLD ROOM	NY	Hold/Staging	228	2025-06-05	40.9
DALLAS F.O. HOLD	TX	Hold/Staging	225	2025-09-18	53.1
MONTGOMERY HOLD RM	TX	Hold/Staging	221	2025-08-21	50.0
BOSTON HOLDROOM	MA	Hold/Staging	206	2025-05-13	18.6
AZ REM OP COORD CENTER (AROCC)	AZ	Hold/Staging	195	2025-04-30	37.4
MIAMI STAGING FACILITY	FL	Hold/Staging	176	2025-01-31	16.5
ATLANTA DIST. HOLD RM	GA	Hold/Staging	168	2025-06-25	42.1
SANTA ANA DRO HOLDROOM	CA	Hold/Staging	150	2025-06-11	11.3
SAN ANTONIO DRO HOLD ROOM	TX	Hold/Staging	145	2025-06-04	19.6
INDIANAPOLIS HOLD ROOM	IN	Hold/Staging	130	2025-09-19	12.3
WASHINGTON FIELD OFFICE	VA	Hold/Staging	127	2025-08-23	17.3
ORLANDO HOLD ROOM	FL	Hold/Staging	126	2025-06-04	19.6

Code

peak_fy25 |>
  filter(peak_pop > 0) |>
  ggplot(aes(x = peak_pop)) +
  geom_histogram(bins = 40) +
  scale_x_log10(labels = comma) +
  labs(x = "Peak detained population (FY25)", y = "Facilities")

Figure 5: Distribution of FY25 peak population among unmatched facilities (log scale, facilities with peak > 0)

Map of unmatched facilities in FY25

Code

map_data <- peak_fy25 |>
  filter(peak_pop > 0, !is.na(latitude), !is.na(longitude))

map_sf <- st_as_sf(map_data, coords = c("longitude", "latitude"), crs = 4326)

type_colors <- c(
  "Dedicated"     = "#e41a1c",
  "Non-Dedicated" = "#377eb8",
  "Federal"       = "#4daf4a",
  "Hold/Staging"  = "#ff7f00",
  "Family/Youth"  = "#984ea3",
  "Medical"       = "#a65628",
  "Other/Unknown" = "#999999",
  "Unclassified"  = "#333333"
)

map_sf$type_grouped <- factor(map_sf$type_grouped, levels = names(type_colors))
present_levels <- levels(droplevels(map_sf$type_grouped))
pal <- colorFactor(palette = type_colors[present_levels], domain = present_levels)

# sqrt scale for area-proportional circle sizing
radii <- scales::rescale(sqrt(map_sf$peak_pop), to = c(3, 20))

map_sf$peak_class <- factor(map_sf$peak_class, levels = c("Peak < 2", "Peak ≥ 2"))

leaflet(map_sf) |>
  fitBounds(lng1 = -125, lat1 = 17.5, lng2 = -65, lat2 = 49.5) |>
  addProviderTiles("CartoDB.Positron") |>

  # Layer 1: Peak < 2  (small solid points, no border)
  addCircleMarkers(
    data        = subset(map_sf, peak_class == "Peak < 2"),
    radius      = 3,                # small fixed dot
    stroke      = FALSE,            # <- no outer border
    fillColor   = ~pal(type_grouped),
    fillOpacity = 1,                # solid
    label       = ~paste0(detention_facility, " (", state, ": peak ", peak_pop, ")"),
    popup       = ~paste0(
      "<b>", detention_facility, "</b><br>",
      state, " | ", type_grouped, "<br>",
      "Peak: ", peak_pop, " (", peak_date, ")<br>",
      "Mean ADP: ", mean_pop
    ),
    group = "Peak < 2"
  ) |>

  # Layer 2: Peak ≥ 2
  addCircleMarkers(
    data        = subset(map_sf, peak_class == "Peak ≥ 2"),
    radius      = radii[map_sf$peak_class == "Peak ≥ 2"],
    color       = "black",
    weight      = 1,
    fillColor   = ~pal(type_grouped),
    fillOpacity = 0.7,
    label       = ~paste0(detention_facility, " (", state, ": peak ", peak_pop, ")"),
    popup       = ~paste0(
      "<b>", detention_facility, "</b><br>",
      state, " | ", type_grouped, "<br>",
      "Peak: ", peak_pop, " (", peak_date, ")<br>",
      "Mean ADP: ", mean_pop
    ),
    group = "Peak ≥ 2"
  ) |>

  # Keep your facility-type legend
  addLegend(
    pal = pal, values = ~type_grouped,
    title = "Facility type", position = "bottomright"
  ) |>

  # Add the layer toggle
  addLayersControl(
    overlayGroups = c("Peak < 2", "Peak ≥ 2"),
    options = layersControlOptions(collapsed = FALSE)
  ) |>
  addSearchFeatures(
    targetGroups = c("Peak < 2", "Peak ≥ 2"),
    options = searchFeaturesOptions(
      zoom = 9,
      openPopup = TRUE,
      position = "topleft",
      hideMarkerOnCollapse = TRUE
    )) |>
  addControl(
    position = "bottomright",
    html = htmltools::tags$style(htmltools::HTML("
      /* Force legend text left in Quarto/Bootstrap contexts */
      .leaflet-container .leaflet-control {
        text-align: left !important;
      }
      .leaflet-container .leaflet-control .legend {
        text-align: left !important;
      }

      /* Make the color swatch and label behave like a left-aligned row */
      .leaflet-container .leaflet-control .legend i {
        display: inline-block !important;
        width: 14px !important;
        height: 14px !important;
        margin: 0 8px 0 0 !important;
        vertical-align: middle !important;
      }
      .leaflet-container .leaflet-control .legend span {
        display: inline !important;
        vertical-align: middle !important;
      }
    ")))

Figure 6: DDP facilities not in ICE FY25 annual statistics, sized by FY25 peak population

Unreported detention: Final Biden quarter vs recent Trump quarter

The preceding sections establish which DDP facilities are absent from the ICE FY25 annual statistics. Here we compare the unreported population—detainees held at those unmatched facilities—during the final three months of the Biden administration versus a comparable window under the Trump administration.

Code

biden_from <- as.Date("2024-10-20")
biden_to   <- as.Date("2025-01-19")
trump_from <- as.Date("2025-07-01")
trump_to   <- as.Date("2025-09-30")

Biden final quarter: 2024-10-20 to 2025-01-19 (92 days)
Trump recent quarter: 2025-07-01 to 2025-09-30 (92 days)

Aggregate unreported population

Code

bind_rows(biden_unmatched, trump_unmatched) |>
  filter(peak_pop > 0) |>
  group_by(period, peak_class) |>
  summarise(
    Facilities = n(),
    `Total ADP` = round(sum(mean_pop)),
    .groups = "drop"
  ) |>
  arrange(peak_class, period) |>
  kable()

Table 5: Unreported DDP population by period (facilities not in ICE FY25)

period	peak_class	Facilities	Total ADP
Biden	Peak < 2	109	8
Trump	Peak < 2	129	14
Biden	Peak ≥ 2	206	396
Trump	Peak ≥ 2	292	3233

Code

bind_rows(biden_unmatched, trump_unmatched) |>
  filter(peak_pop > 0, peak_class == "Peak \u2265 2") |>
  group_by(period) |>
  summarise(
    Facilities = n(),
    `Total ADP` = round(sum(mean_pop)),
    `Mean ADP per facility` = round(sum(mean_pop) / n(), 1),
    .groups = "drop"
  ) |>
  kable()

Table 6: Unreported population totals (Peak ≥ 2 facilities only)

period	Facilities	Total ADP	Mean ADP per facility
Biden	206	396	1.9
Trump	292	3233	11.1

This graph shows the distribution of population totals among the unmatched facilities in each period, with a horizontal line marking the average for each period. The six largest facilities are named. As you can see, the largest unreported opulation in the Biden period would not even be an outlier in the Trump period. Conversely, the three largest unreported facilities in the Trump period each held more people in this period that all the Biden facilities combined.

Code

source("violin-plot.R")
violin_plot

By facility type

Code

type_comp <- bind_rows(biden_unmatched, trump_unmatched) |>
  filter(peak_pop > 0, adp_class == "ADP \u2265 2") |>
  group_by(period, type_grouped) |>
  summarise(
    Facilities = n(),
    ADP = round(sum(mean_pop)),
    .groups = "drop"
  )

type_comp |>
  pivot_wider(names_from = period,
              values_from = c(Facilities, ADP),
              values_fill = 0,
              names_glue = "{period} {.value}") |>
  mutate(
    `ADP change` = `Trump ADP` - `Biden ADP`,
    `% change` = if_else(
      `Biden ADP` == 0, NA_real_,
      round((`Trump ADP` / `Biden ADP` - 1) * 100, 1)
    )
  ) |>
  arrange(desc(`Trump ADP`)) |>
  kable(col.names = c("Facility type",
                       "Biden fac.", "Trump fac.",
                       "Biden ADP", "Trump ADP",
                       "ADP change", "% change"))

Table 7: Unreported facilities by type and period (ADP ≥ 2 only)

Facility type	Biden fac.	Trump fac.	Biden ADP	Trump ADP	ADP change	% change
Hold/Staging	37	91	258	1848	1590	616.3
Dedicated	0	1	0	808	808	NA
Unclassified	8	20	32	348	316	987.5
Federal	1	11	7	71	64	914.3
Non-Dedicated	2	10	14	48	34	242.9

Code

bind_rows(biden_unmatched, trump_unmatched) |>
  filter(peak_pop > 0, peak_class == "Peak < 2") |>
  group_by(period, type_grouped) |>
  summarise(
    Facilities = n(),
    ADP = round(sum(mean_pop), 1),
    .groups = "drop"
  ) |>
  pivot_wider(names_from = period,
              values_from = c(Facilities, ADP),
              values_fill = 0,
              names_glue = "{period} {.value}") |>
  arrange(desc(`Trump Facilities`)) |>
  kable(col.names = c("Facility type",
                       "Biden fac.", "Trump fac.",
                       "Biden ADP", "Trump ADP"))

Table 8: Transient facilities (Peak < 2) by type and period

Facility type	Biden fac.	Trump fac.	Biden ADP	Trump ADP
Unclassified	77	105	5.4	11.6
Federal	4	9	0.2	0.3
Non-Dedicated	9	8	1.3	2.1
Hold/Staging	17	5	0.5	0.2
Medical	2	1	0.1	0.1
Family/Youth	0	1	0.0	0.0

Facilities active in only one period

Code

biden_active <- biden_unmatched |> filter(peak_pop > 0, peak_class == "Peak \u2265 2")
trump_active <- trump_unmatched |> filter(peak_pop > 0, peak_class == "Peak \u2265 2")

biden_only_codes <- setdiff(biden_active$detention_facility_code,
                             trump_active$detention_facility_code)
trump_only_codes <- setdiff(trump_active$detention_facility_code,
                             biden_active$detention_facility_code)

Code

if (length(biden_only_codes) > 0) {
  biden_unmatched |>
    filter(detention_facility_code %in% biden_only_codes) |>
    arrange(desc(mean_pop)) |>
    select(detention_facility, state, type_grouped, mean_pop, peak_pop) |>
    mutate(mean_pop = round(mean_pop, 1)) |>
    kable(col.names = c("Facility", "State", "Type", "Biden ADP", "Biden peak")) 
}

Table 9: Unreported facilities with Peak ≥ 2 active under Biden but not Trump

Facility	State	Type	Biden ADP	Biden peak
NYE COUNTY SHERIFF-PAHRUMP	NV	Non-Dedicated	12.0	69
GUAYNABO MDC (SAN JUAN)	PR	Federal	6.6	12
MONTGOMERY CO MAC SIM BUTLER D.F.	AL	Federal	0.9	2
BRONSON METHODIST HOSPITAL	MI	Unclassified	0.8	2
COBB CO SHERIFFS OFC ADULT DET CTR	GA	Non-Dedicated	0.6	4
SAINT ALBANS HOLD ROOM	VT	Hold/Staging	0.5	7
LINCOLN COUNTY SHERIFF’S	MO	Non-Dedicated	0.4	3
ST. JOHN’S COUNTY JAIL	FL	Unclassified	0.4	2
METHODIST HOSPITAL TEXSAN	TX	Unclassified	0.4	2
WASHINGTON COUNTY JAIL	MN	Unclassified	0.4	3
ORANGE REGIONAL MED CTR	NY	Unclassified	0.2	2
LOMPOC FCI	CA	Unclassified	0.2	6
COLUMBUS HOLDROOM	OH	Hold/Staging	0.2	9
KARNES COUNTY CIVIL DET. FACILITY	TX	Family/Youth	0.1	2
KINNEY COUNTY JAIL	TX	Unclassified	0.1	4
SIOUX CITY HOLD ROOM	IA	Hold/Staging	0.1	3
EL CENTRO REGIONAL MED CTR	CA	Unclassified	0.1	2
METHODIST HOSPITAL	TX	Unclassified	0.1	2
SPRINGFIELD FED.MED.CNTR	MO	Unclassified	0.1	2
FCI VICTORVILLE	CA	Unclassified	0.1	7
ST FRANCIS HOSPITAL	GA	Unclassified	0.1	2
CHANDLER REG MED CTR	AZ	Unclassified	0.1	2
CINCINNATI HOLDROOM	OH	Hold/Staging	0.0	2
SNJ HOLD ROOM	CA	Hold/Staging	0.0	2

24 unreported facilities (Peak ≥ 2) were active in the Biden period but fell below the threshold or had zero population in the Trump period. Of these, 2 had ADP ≥ 2 under Biden, contributing 19 ADP in total.

Code

if (length(trump_only_codes) > 0) {
  trump_unmatched |>
    filter(detention_facility_code %in% trump_only_codes) |>
    arrange(desc(mean_pop)) |>
    select(detention_facility, state, type_grouped, mean_pop, peak_pop) |>
    mutate(mean_pop = round(mean_pop, 1)) |>
    kable(col.names = c("Facility", "State", "Type", "Trump ADP", "Trump peak")) |> 
    scroll_box(height = "630px")
}

Table 10: Unreported facilities with Peak ≥ 2 active under Trump but not Biden (scroll for full list)

Facility	State	Type	Trump ADP	Trump peak
FLORIDA SOFT-SIDED FACILITY-SOUTH	FL	Dedicated	808.4	1482
EL PASO SOFT SIDED FACILITY	TX	Hold/Staging	326.1	1606
BAKER C. I.	FL	Unclassified	176.1	667
URSULA CENTRALIZED PROCESSING CNTR	TX	Unclassified	84.6	365
BROOKS COUNTY JAIL (CONTRACT)	TX	Federal	21.1	90
TURNER GUIFORD KNIGHT (TGK) JAIL	FL	Non-Dedicated	14.9	27
WESTERN TENNESSEE DET. FAC.	TN	Federal	12.6	74
MARTIN COUNTY JAIL	FL	Non-Dedicated	7.4	24
CHRISTIAN COUNTY JAIL	KY	Federal	6.4	51
WESTMINSTER CUSTODY	CA	Hold/Staging	6.0	7
RIO VISTA BEHV HLTH HSP	TX	Unclassified	5.2	14
ERO HOLD ROOM (SPRINGFIELD MO)	MO	Hold/Staging	5.0	62
OSCEOLA COUNTY JAIL	FL	Unclassified	4.5	12
EASTERN REGIONAL JAIL	WV	Federal	4.5	23
SW VIRGINIA REG JAIL AUTH-ABG FAC	VA	Unclassified	4.0	47
SIOUX COUNTY JAIL	IA	Federal	4.0	11
CHAVEZ DET CRT	NM	Non-Dedicated	3.5	16
TOM GREEN CTY JAIL	TX	Federal	3.2	11
LAWRENCE CO. JAIL, SD	SD	Non-Dedicated	2.7	11
GRAND RAPIDS HOLDROOM	MI	Hold/Staging	2.7	9
NEW YORK PRESBYTERIAN HOSPITAL	NY	Unclassified	2.4	5
CHESAPEAKE CITY JAIL	VA	Unclassified	2.3	25
CEDAR RAPIDS HOLD ROOM	IA	Hold/Staging	2.2	17
CASCADE COUNTY JAIL, MT	MT	Federal	2.1	5
HILLSBOROUGH COUNTY JAIL	FL	Non-Dedicated	2.0	12
WHITE MEMORIAL HOSPITAL	CA	Unclassified	1.8	4
SCOTT COUNTY JAIL	IA	Unclassified	1.8	7
SYRACUSE HOLD ROOM	NY	Hold/Staging	1.8	14
SEMINOLE COUNTY JAIL	FL	Unclassified	1.7	5
CACHE CO. JAIL	UT	Federal	1.7	13
BUFFALO HOLD ROOM	NY	Hold/Staging	1.7	17
ST JOSEPH MEDICAL CENTER	WA	Unclassified	1.7	4
MINICASSIA DET. CENTER	ID	Non-Dedicated	1.6	7
LARAMIE COUNTY JAIL	WY	Federal	1.6	8
RIO GRANDE VALLEY STAGING	TX	Hold/Staging	1.5	11
MARION COUNTY JAIL	FL	Unclassified	1.5	5
COLO SPRINGS DEN HSI HOLD	CO	Hold/Staging	1.5	9
LIRS - LFSRM FORT COLLINS LTFC CO	CO	Unclassified	1.5	2
TEX HEALTH HUGULEY HOSP	TX	Medical	1.3	4
NASSAU COUNTY ICE TRAILER	NY	Unclassified	1.3	13
UNIVERSITY MED CENTER OF EL PASO	TX	Unclassified	1.2	3
BREVARD COUNTY JAIL COMPLEX	FL	Unclassified	1.2	5
AGUADILLA SPC	PR	Hold/Staging	1.2	26
EL PASO LTAC HOSPITAL	TX	Unclassified	1.1	3
PROMISE HOSPITAL	AZ	Unclassified	1.1	3
ALVARADO PARKWAY INSTITUTE	CA	Unclassified	1.0	3
UCHEALTH UNV HOSP OF CO	CO	Unclassified	0.9	2
ANAHEIM GLOBAL MEDICAL CENTER	CA	Unclassified	0.9	3
MANATEE COUNTY DETENTION-ANNEX	FL	Unclassified	0.9	7
VRK PRCS	NY	Hold/Staging	0.9	12
HARRIS COUNTY PSYCHIATRIC CENTER	TX	Unclassified	0.9	2
DADE CORRECTIONAL INST	FL	Dedicated	0.8	11
BUFFALO USBP HOLD ROOM	NY	Federal	0.7	20
POLK COUNTY JAIL	FL	Unclassified	0.7	23
LOUISVILLE HOLD ROOM	KY	Hold/Staging	0.7	4
CBP MOVEMENT COORDINATION AREA	NA	Unclassified	0.7	12
FRANKLIN COUNTY JAIL	PA	Federal	0.6	3
BROWN COUNTY JAIL	WI	Unclassified	0.6	4
CHAMPLAIN HOLD ROOM	NY	Hold/Staging	0.5	8
HOUSTON HEALTHCARE HOSP	TX	Unclassified	0.5	3
LOUDOUN COUNTY	VA	Unclassified	0.5	4
NORTHWEST REG. ADULT DET.	VA	Unclassified	0.5	3
WENTWORTH DOUGLAS HOSPITAL	NH	Unclassified	0.5	2
RAPID CITY HOLD ROOM	SD	Hold/Staging	0.5	7
FRIO REGIONAL HOSPITAL	TX	Unclassified	0.5	2
CLEVELAND HOLD ROOM	OH	Hold/Staging	0.4	14
NASHVILLE HOLD ROOM	TN	Hold/Staging	0.4	22
BROWARD GENERAL MEDICAL CENTER	FL	Unclassified	0.4	3
UNION COUNTY JAIL	SD	Unclassified	0.4	9
GLENWOOD SPRINGS HOLDROOM	CO	Hold/Staging	0.4	3
MEDFORD SUB-OFFICE	OR	Hold/Staging	0.4	14
UNIVERSITY HOSPITAL	NJ	Unclassified	0.4	2
ALAMOSA HOLDROOM	CO	Hold/Staging	0.4	7
EUGENE HOLDROOM	OR	Hold/Staging	0.4	7
MONROE COUNTY JAIL	FL	Non-Dedicated	0.4	2
LOYOLA CENTER FOR HEALTH	IL	Unclassified	0.4	6
TX HLTH HARRIS METH HOSP	TX	Unclassified	0.4	2
LAREDO MEDICAL CENTER	TX	Unclassified	0.3	2
NASSAU UNIV MED CTR	NY	Unclassified	0.3	3
GRADY MEMORIAL HOSPITAL	GA	Unclassified	0.3	2
HIGHLAND MTN COR CTR	AK	Unclassified	0.3	2
DURANGO HOLDROOM	CO	Hold/Staging	0.3	6
ARROWHEAD REGIONAL MED CENTER	CA	Unclassified	0.3	3
MCLENNAN COUNTY JAIL	TX	Unclassified	0.3	2
BOSTON SPC	MA	Unclassified	0.3	2
CENTRAL VA REGIONAL JAIL	VA	Unclassified	0.3	2
VIRGINIA BEACH	VA	Unclassified	0.3	2
CAMPBELL COUNTY JAIL	WY	Unclassified	0.3	4
FLAGLER COUNTY JAIL	FL	Unclassified	0.3	12
SUMTER COUNTY JAIL	FL	Unclassified	0.3	2
BANNER IRONWOOD MED CTR	AZ	Unclassified	0.2	2
CRAIG HOLDROOM	CO	Hold/Staging	0.2	4
PARKLAND HOSPITAL	TX	Unclassified	0.2	3
WEST MIAMI DADE BAPT HEALTH EMERG	FL	Unclassified	0.2	2
SALINE COUNTY JAIL	NE	Non-Dedicated	0.2	7
ST ELIZABETH YOUNGSTOWN HOSPITAL	OH	Unclassified	0.2	2
UNIVERSITY MEDICAL CENTER	NV	Unclassified	0.2	2
CHARLESTON COUNTY CORRECT	SC	Federal	0.1	3
EL PASO COUNTY JAIL (TX)	TX	Unclassified	0.1	2
PALM BEACH COUNTY JAIL	FL	Unclassified	0.1	4
MOUNT NITTANY MEDICAL CENTER	PA	Unclassified	0.1	2
RIO GRANDE ST.HOSPITAL	TX	Unclassified	0.1	7
CHARLOTTE COUNTY JAIL	FL	Unclassified	0.1	2
ANCHORAGE SUB-OFFICE HOLD ROOM	AK	Hold/Staging	0.1	5
GARDEN CITY PROCESSING @ FINNEY CO	KS	Unclassified	0.1	3
HEL DISTRICT CUSTODY	MT	Hold/Staging	0.1	2
FORT SMITH HOLDROOM	AR	Hold/Staging	0.0	3
HENDERSONVILLE HOLD ROOM	NC	Hold/Staging	0.0	3
WICHITA HOLDROOM	KS	Hold/Staging	0.0	2
MARION CO JAIL	OR	Unclassified	0.0	2

110 unreported facilities (Peak ≥ 2) were active in the Trump period but not the Biden period. Of these, 25 had ADP < 2 under Trump, contributing 1514 ADP in total.

Code

if (length(trump_only_codes) > 0) {
  trump_unmatched |>
    filter(detention_facility_code %in% trump_only_codes) |>
    group_by(type_grouped) |>
    summarise(
      Facilities = n(),
      `Total ADP` = round(sum(mean_pop)),
      .groups = "drop"
    ) |>
    arrange(desc(`Total ADP`)) |>
    kable(col.names = c("Facility type", "Facilities", "Total ADP"))
}

Table 11: New unreported facilities under Trump (ADP ≥ 2), by type

Facility type	Facilities	Total ADP
Dedicated	2	809
Hold/Staging	27	355
Unclassified	60	311
Federal	12	59
Non-Dedicated	8	33
Medical	1	1

Facilities active in both periods

Code

both_codes <- intersect(biden_active$detention_facility_code,
                         trump_active$detention_facility_code)

both_compare <- biden_unmatched |>
  filter(detention_facility_code %in% both_codes) |>
  select(detention_facility_code, detention_facility, state, type_grouped,
         biden_adp = mean_pop, biden_peak = peak_pop) |>
  inner_join(
    trump_unmatched |>
      filter(detention_facility_code %in% both_codes) |>
      select(detention_facility_code,
             trump_adp = mean_pop, trump_peak = peak_pop),
    by = "detention_facility_code"
  ) |>
  mutate(
    adp_change = trump_adp - biden_adp,
    adp_pct_change = if_else(biden_adp == 0, NA_real_,
                             (trump_adp / biden_adp - 1) * 100)
  )

Code

ggplot(both_compare, aes(x = biden_adp, y = trump_adp, color = type_grouped)) +
  geom_abline(slope = 1, intercept = 0, linetype = "dashed", alpha = 0.5) +
  geom_point(alpha = 0.7) +
  scale_x_log10(labels = comma) +
  scale_y_log10(labels = comma) +
  labs(x = "Biden final quarter ADP", y = "Trump recent quarter ADP",
       color = "Facility type")

Figure 7: Mean ADP at unreported facilities: Biden vs Trump (facilities with ADP ≥ 2 in both periods, log scale)

Code

both_compare |>
  arrange(desc(adp_change)) |>
  head(15) |>
  transmute(
    Facility = detention_facility, State = state, Type = type_grouped,
    `Biden ADP` = round(biden_adp),
    `Trump ADP` = round(trump_adp),
    Change = round(adp_change)
  ) |>
  kable()

Table 12: Top 15 unreported facilities by ADP increase (Biden → Trump)

Facility	State	Type	Biden ADP	Trump ADP	Change
DALLAS F.O. HOLD	TX	Hold/Staging	16	104	88
LOS CUST CASE	CA	Hold/Staging	7	89	82
ATLANTA DIST. HOLD RM	GA	Hold/Staging	9	89	80
BROADVIEW SERVICE STAGING	IL	Hold/Staging	9	89	80
MONTGOMERY HOLD RM	TX	Hold/Staging	30	96	66
NYC HOLD ROOM	NY	Hold/Staging	5	67	62
AZ REM OP COORD CENTER (AROCC)	AZ	Hold/Staging	16	70	54
BALTIMORE HOLD ROOM	MD	Hold/Staging	2	47	45
WASHINGTON FIELD OFFICE	VA	Hold/Staging	7	44	37
MIAMI (MIRAMAR) HOLD ROOM	FL	Hold/Staging	0	35	35
SAN ANTONIO DRO HOLD ROOM	TX	Hold/Staging	5	36	31
RICHMOND SUB-OFFICE HOLD	VA	Hold/Staging	5	34	29
ORLANDO HOLD ROOM	FL	Hold/Staging	6	34	28
TAMPA HOLD ROOM	FL	Hold/Staging	5	33	27
SND DISTRICT STAGING	CA	Hold/Staging	9	35	27

Code

both_compare |>
  arrange(adp_change) |>
  head(15) |>
  transmute(
    Facility = detention_facility, State = state, Type = type_grouped,
    `Biden ADP` = round(biden_adp),
    `Trump ADP` = round(trump_adp),
    Change = round(adp_change)
  ) |>
  kable()

Table 13: Top 15 unreported facilities by ADP decrease (Biden → Trump)

Facility	State	Type	Biden ADP	Trump ADP	Change
KROME HOLD ROOM	FL	Hold/Staging	21	0	-21
MCALLEN HOLD ROOM	TX	Hold/Staging	13	5	-7
PECOS HOLD ROOM	TX	Hold/Staging	3	1	-1
COLUMBIA CARE CENTER	SC	Unclassified	3	2	-1
NEW/INS OS HOLD ROOM	NJ	Hold/Staging	1	0	-1
CASPER HOLDROOM	WY	Hold/Staging	0	0	0
CHARLOTTE AMALIE HOLDROOM	VI	Hold/Staging	0	0	0
BALDWIN COUNTY COR. CENTER	AL	Non-Dedicated	2	2	0
FAYETTEVILLE HOLD ROOM	AR	Hold/Staging	0	0	0
LAS CRUCES HOLD ROOM	NM	Hold/Staging	0	0	0
JACK HARWELL DETENTION CENTER	TX	Federal	0	0	0
VALLEYWISE HEALTH MED CTR	AZ	Unclassified	0	0	0
LEXINGTON PROCESSING @ FAYETTE CO	KY	Unclassified	0	0	0
ROCK ISLAND HOLD ROOM	IL	Hold/Staging	1	1	0
TITUS COUNTY JAIL	TX	Federal	0	0	0

Daily unreported population by period

Code

ggplot(daily_unreported, aes(x = date, y = total_pop, fill = type_grouped)) +
  geom_col() +
  facet_wrap(~period, scales = "free_x") +
  scale_y_continuous(labels = comma) +
  scale_x_date(date_labels = "%b %d") +
  labs(x = NULL, y = "Unreported detained population", fill = "Facility type")

Figure 8: Daily unreported detention population during both comparison periods, by facility type (ADP ≥ 2 facilities only)