How lawyer-record comparisons are calculated
See how qualifying representative filing records are linked to historical asylum outcomes and which comparisons this proxy supports.
What is counted
The unchanged M1 denominator contains one eligible proceeding event with a nonconflicting ASYL grant (G/I/P) or denial (D), a consistent case join and proceeding completion in FY2021–FY2025. G/(G+D) is calculated within each group; Wilson 95% intervals use the same G and N.
Qualifying record found
An eligible court/respondent tbl_RepsAssigned row must be parented to A_TblCase, have matching positive case references and a matching accepted case record. We take the earliest retained E_28_DATE between 1980-01-01 and 2026-08-31, using calendar dates. A qualifying match exists when that date is before or on proceeding completion and the case has no conflicting assignment references.
No qualifying record found
This is the complementary result of the same record search, including no retained row, a first valid date after completion, no usable date, or unresolved case linkage. These records stay in M1. It is not a category of proven unrepresented people. The original research on active representation remains valid: the export lacks scope, acceptance and complete relationship-end history.
Interpretation
The displayed gap is the found-proxy grant share minus the not-found-proxy grant share, in percentage points, whenever both groups have a nonzero eligible denominator. It is retained even when either group has N < 100; that group and the gap receive a small-sample cue, not a new suppression rule. A filing date does not prove that the representative was active or covered the asylum decision. Case-linked dates may span earlier proceedings, and same-day ordering is unknown. The groups may differ in legal merits, detention, nationality and other characteristics. This is a descriptive historical difference, not a causal effect or individual prediction.
Publication and disclosure
The historical RP1 release masked representation-record pairs using sample thresholds and subtraction checks. Its original files remain available as historical artifacts. The current site uses a complete public-source edition for every observed national, court, judge and nationality intersection. Every pair reconciles to the same public grant-and-denial denominator; no local small-cell mask remains on these proxy counts. A zero-record side has no grant share.
Source and calculation details
The source code key distinguishes filing dates, proceeding completion, representative roles and case linkage. “Representative” can include a qualified non-lawyer, so the product does not equate every EOIR-28 with an attorney.