The Honorable Immigration Judge Raisa Cohen earned a Bachelor of Business Administration in 2002 from Baruch College's Zicklin School of Business and a Juris Doctor in 2007 from St. John's University School of Law. She served as an assistant chief counsel for U.S. Immigration and Customs Enforcement from April 2009 to September 2014 and again from September 2015 to February 2016. Between those roles, she was an attorney at Cohen & Cohen Law Group PC from October 2014 to September 2015, and earlier worked as an immigration attorney at the Law Firm of Ted Sofer from 2007 to 2009. Judge Cohen is a member of the New York State Bar.
| Asylum | 92% | n=448 High confidence | |
| Withholding / CAT | 40% | n=75 Moderate | |
| Cancellation — 42B (non-LPR) | 95.6% | n=68 Moderate | |
| Cancellation — 42A (LPR) | 100%⚠ | n=5 Too few cases | |
| Adjustment of status (245) | 98.5% | n=67 Moderate | |
| Voluntary departure | 100%⚠ | n=4 Too few cases |
Source. U.S. Department of Justice EOIR immigration court records, released under FOIA (data through 2026-06-01), packaged as a public DuckDB database. The same underlying records used by TRAC and the Deportation Data Project. Judge appointment dates and appointing Attorneys General are compiled separately from EOIR/DOJ biographies.
Asylum grant rate. Grants ÷ (grants + denials) among asylum applications decided on the merits, for decisions completed since 2021. Cases that were withdrawn, abandoned, administratively closed, or not adjudicated are excluded.
Bond decisions. From EOIR's bond (custody redetermination) records since 2021. "Granted" = bond amount decreased, release on own recognizance, or amelioration granted; "denied" = no change, no bond set, amelioration denied, or denied as moot. Withdrawn requests, jurisdictional denials, and bond increases are excluded from the grant rate. Decision codes follow the Deportation Data Project codebook; EOIR's own DHS "initial bond" field is unreliably filled and is not used. The median bond reflects decisions where the judge set or adjusted a dollar amount.
Representation. A person is counted as "represented" if a respondent's attorney (EOIR-28) appeared at the court level.
The map & appointing party. Each state is shaded by the mix of appointing administrations of the immigration judges currently sitting at the court(s) covering that state — blue where more were appointed under Democratic administrations, red where more under Republican ones. States with no immigration court of their own are drawn with diagonal stripes and take the mix of the court(s) with administrative control over them, per EOIR's administrative control list (e.g., Montana, Idaho, and Alaska → Portland, OR; Wyoming → Denver; the Dakotas → Fort Snelling, MN; Mississippi is split between Memphis and New Orleans). EOIR defines control by DHS district offices and detention facilities rather than clean geographic lines, so in states with several courts the map does not draw internal boundaries; instead the hover and state pages break the numbers down court by court. Where the control list assigns dockets by custody status, each court is labeled non-detained, detained (courts sitting at or primarily serving detention facilities and Service Processing Centers), or all cases; which court hears a given case depends on custody status and where DHS filed it. Party is taken from the appointing Attorney General (or President). This is emphatically not a measure of how judges rule. Appointing administration is a weak-to-nonexistent predictor of individual case outcomes, and judges appointed by any administration span the full range of grant rates. The map is offered only to show the composition of the bench, not to forecast decisions.
Courts where a judge decides & WebEx links. The list of a judge's hearing locations counts the courts (EOIR base city) on that judge's completed cases in the period. Internet-based hearing (WebEx) links and telephonic access codes are reproduced from EOIR's public Find an Immigration Court page and apply only to parties with a scheduled hearing.
Guardrails. Every figure shows its sample size and a confidence label (High ≥200, Moderate 50–199). A rate based on fewer than 50 merits decisions is still shown but marked with a ⚠ icon meaning it is statistically unreliable and should not be read as a stable pattern.