diff --git a/DESCRIPTION b/DESCRIPTION
index 7159bab..ec395ac 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -1,7 +1,7 @@
 Package: uscogdata
 Type: Package
 Title: Curated Reader for the Civilytics US Census of Governments Finance Corpus
-Version: 0.3.0
+Version: 0.4.0
 Authors@R: c(
     person(c("Jared", "E."), "Knowles",
            email = "jared@civilytics.com",
diff --git a/NEWS.md b/NEWS.md
index cace5cd..6c3b108 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,3 +1,59 @@
+# uscogdata 0.4.0
+
+## Cohorts can be named by predicate, not just by id
+
+`cog_spending()`, `cog_revenue()` and `cog_balances()` gain optional `state`
+and `type` arguments. Both default to `NULL`, so every existing call behaves
+exactly as before.
+
+Passing them expresses the cohort as a subquery against `canonical_fips_xwalk`
+inside each statement, instead of round-tripping the ids through R and
+rendering them back into a literal `IN` list:
+
+```r
+# before: resolve 20,106 ids in R, then embed them in every statement
+ids <- cog_gov_search(NULL, state = "CA", type = "city")$canonical_govid
+cog_spending(ids, years = 2022)
+
+# now: the cohort never leaves the database
+cog_spending(years = 2022, state = "CA", type = "city")
+```
+
+Measured against the production corpus, same FY2022 aggregate over the
+20,106-government `type = "city"` cohort:
+
+| cohort expressed as | time |
+|---|---:|
+| `IN (20,106 literals)` | 449 ms |
+| join against a temp cohort table | 99 ms |
+| predicate on `canonical_fips_xwalk` | **94 ms** |
+| no cohort filter at all (the floor) | 88 ms |
+
+**4.8x, within 7% of the floor.** The rendered `IN` list was 301,591
+characters and was re-parsed in 5-8 separate statements per call, so the cost
+was paid repeatedly; the predicate's size is constant in the cohort.
+
+`state` and `type` use the same vocabulary and the same internal coercion as
+`cog_gov_search()` -- `state` is a postal abbreviation (`"WI"`) even though the
+crosswalk column holds a FIPS code (`"55"`).
+
+Supplying `govid` **and** `state`/`type` intersects them: the governments in
+`govid` that also match the predicate. Naming no cohort at all now aborts with
+class `uscogdata_no_cohort` rather than R's "argument is missing" error.
+
+When the cohort is named by predicate there is no id list to report, so
+`provenance$scope$govids_found`/`govids_missing` are empty and
+`provenance$scope$cohort` carries `state`, `type` and `n_governments` instead.
+A `govid`-named cohort's provenance is unchanged.
+
+## Fixes
+
+* An unknown `state` abbreviation now aborts with "Unknown state abbreviation"
+  (class `uscogdata_unknown_state`) instead of base R's "subscript out of
+  bounds". `.state_abbrev_to_fips` is a named character vector, so `[[` on an
+  absent name threw before the curated message could be reached -- making that
+  message unreachable dead code in every verb that takes a `state`.
+
 # uscogdata 0.3.0
 
 First public release.
diff --git a/R/balances.R b/R/balances.R
index 994e0b2..26b439c 100644
--- a/R/balances.R
+++ b/R/balances.R
@@ -18,7 +18,9 @@
 #' comparable to a GAAP fund balance from an ACFR.
 #'
 #' @param govid Canonical govid(s): a character vector, or a data frame with a
-#'   `canonical_govid` column (e.g. from [cog_gov_search()]).
+#'   `canonical_govid` column (e.g. from [cog_gov_search()]). `NULL` to name
+#'   the cohort by `state`/`type` instead.
+#' @inheritParams cog_spending
 #' @param years Integer vector of fiscal years.
 #' @param category Optional character vector of categories to keep. One of
 #'   `"Fund Balances"`, `"Insurance Trust Balances"`,
@@ -63,15 +65,16 @@
 #'   requested years). `expenditure_concept`/`revenue_concept` are `NA` --
 #'   holdings are a stock, not a flow, so neither concept vocabulary applies.
 #' @export
-cog_balances <- function(govid, years, category = NULL,
+cog_balances <- function(govid = NULL, years, category = NULL,
                          per_capita = FALSE, adjust_to_year = NULL,
-                         basis = c("harmonized", "raw"), recipe = NULL) {
+                         basis = c("harmonized", "raw"), recipe = NULL,
+                         state = NULL, type = NULL) {
   call <- match.call()
   basis <- match.arg(basis, c("harmonized", "raw"))
   # Coerce FIRST, validate second: .validate_verb_inputs() asserts
   # is.character(govid), and a data-frame govid (cog_gov_search() output) has
   # not been unwrapped yet at this point.
-  govid <- .coerce_govid_input(govid)
+  govid <- if (is.null(govid)) NULL else .coerce_govid_input(govid)
   # The money verbs' validator, reused rather than re-implemented (R/spending.R).
   # It covers the exact superset cog_balances() needs -- including the
   # recipe/category mutual-exclusivity guard -- so a second local copy would
@@ -91,6 +94,8 @@ cog_balances <- function(govid, years, category = NULL,
   years <- as.integer(years)
   if (!is.null(adjust_to_year)) adjust_to_year <- as.integer(adjust_to_year)
 
+  cohort <- .make_cohort(govid, state, type)
+
   con <- .ensure_session()
   .require_balance_support(con)
   scope <- .check_govids_in_scope(govid)
@@ -109,7 +114,7 @@ cog_balances <- function(govid, years, category = NULL,
     .validate_recipe_id(con, recipe)
     comps <- .recipe_components(con, recipe)
     recipe_label <- comps$label[[1]]
-    result <- .run_recipe(con, recipe, govid, years)
+    result <- .run_recipe(con, recipe, cohort, years)
     sql <- attr(result, "sql_query")
     result <- .shape_recipe_result(result, "balance_subtype", recipe_label)
     recipe_block <- list(
@@ -119,7 +124,7 @@ cog_balances <- function(govid, years, category = NULL,
     category_for_prov <- recipe_label
   } else {
     sql <- .build_verb_sql("balance_annotated", "balance_subtype",
-                           govid, years, category,
+                           cohort, years, category,
                            ig_view = NULL, subtype_scope = NULL)
     result <- tibble::as_tibble(DBI::dbGetQuery(con, sql))
   }
@@ -127,7 +132,7 @@ cog_balances <- function(govid, years, category = NULL,
   # Order matters (matches .verb_spendrev()): per-capita first, so
   # .attach_real_dollars() deflates the nominal per-capita column into
   # amt_per_capita_real rather than needing amt_per_capita_nominal recomputed.
-  if (isTRUE(per_capita)) result <- .attach_per_capita(result, con, govid)
+  if (isTRUE(per_capita)) result <- .attach_per_capita(result, con)
   if (!is.null(adjust_to_year)) {
     result <- .attach_real_dollars(result, adjust_to_year, per_capita)
   }
@@ -145,6 +150,7 @@ cog_balances <- function(govid, years, category = NULL,
   )
   prov$scope$govids_found   <- scope$found
   prov$scope$govids_missing <- scope$missing
+  prov$scope$cohort <- .cohort_provenance(con, cohort)
 
   prov$balance_caveats <- .balance_caveats(
     con, prov$codes_summed$observed, years
diff --git a/R/basis.R b/R/basis.R
index dd60a67..b173cbb 100644
--- a/R/basis.R
+++ b/R/basis.R
@@ -54,7 +54,7 @@
 #' expenditure_concept = "total": ig_long_harmonized COALESCEs rather than
 #' drops NULL-harmonized rows, so harmonization never excludes an IG row.
 #' @noRd
-.build_harmonization_block <- function(con, govid, years, resolved,
+.build_harmonization_block <- function(con, cohort, years, resolved,
                                        subtype_col, subtype_scope) {
   if (!identical(resolved$basis, "harmonized")) {
     return(list(
@@ -68,12 +68,12 @@
   sql <- sprintf(
     "SELECT COUNT(*) AS n, COALESCE(SUM(amt), 0) * 1000.0 AS amt
      FROM long
-     WHERE canonical_govid IN (%s) AND year IN (%s)
+     WHERE %s AND year IN (%s)
        AND NOT is_aggregate AND harmonized_code IS NULL
        AND item_code IN (
          SELECT item_code FROM summary_categories WHERE %s IN (%s)
        )",
-    .sql_lit_chr(govid), paste(as.integer(years), collapse = ","),
+    .cohort_sql(cohort), paste(as.integer(years), collapse = ","),
     subtype_col, .sql_lit_chr(subtype_scope)
   )
   na <- DBI::dbGetQuery(con, sql)
diff --git a/R/cohort.R b/R/cohort.R
new file mode 100644
index 0000000..67fc16f
--- /dev/null
+++ b/R/cohort.R
@@ -0,0 +1,127 @@
+# How a verb names the set of governments it queries.
+#
+# Historically there was one way: a `govid` character vector, rendered by
+# .sql_lit_chr() into a quoted IN list. That is fine for a handful of
+# governments and pathological for a fleet. Measured against the production
+# corpus, the same FY2022 aggregate over the 20,106-government `type = "city"`
+# cohort:
+#
+#   cohort expressed as                            time
+#   IN (20,106 literals)                           449 ms
+#   join against a temp cohort table                99 ms
+#   predicate on canonical_fips_xwalk               94 ms
+#   no cohort filter at all (the floor)             88 ms
+#
+# 4.8x, and within 7% of the no-filter floor. The rendered IN list is 301,591
+# characters and .verb_spendrev() embeds it in 5-8 separate statements per
+# call, so the parse-and-plan cost is paid over and over (uscogdata#58).
+#
+# A cohort therefore has two independent halves, and a query can carry either
+# or both:
+#
+#   ids        an explicit canonical_govid vector      -> literal IN list
+#   predicate  state/type over canonical_fips_xwalk    -> IN (SELECT ...)
+#
+# Both together is an INTERSECTION -- "these ids, narrowed to that state/type"
+# -- never a precedence rule where one silently wins.
+
+#' Build the internal cohort object shared by every query verb.
+#'
+#' `state` and `type` are coerced with the SAME helpers `cog_gov_search()`
+#' uses. That is load-bearing, not tidiness: the public argument is a postal
+#' abbreviation (`"WI"`) while `canonical_fips_xwalk.fips_state` holds a FIPS
+#' code (`"55"`), and `type` is a label (`"city"`) against an integer
+#' `govs_type`. A predicate written against the raw parameter matches nothing
+#' and returns an empty result indistinguishable from "this government
+#' reported nothing" -- cog-api hit exactly that trap optimizing this path.
+#' One definition of the translation, not two.
+#'
+#' @param govid Already-coerced character vector of canonical_govids, or NULL.
+#' @param state Postal abbreviation or FIPS code, or NULL.
+#' @param type Type label or integer code, or NULL.
+#' @noRd
+.make_cohort <- function(govid = NULL, state = NULL, type = NULL) {
+  if (is.null(govid) && is.null(state) && is.null(type)) {
+    cli::cli_abort(c(
+      "A cohort must be named.",
+      "*" = "Pass {.arg govid} for specific governments, or {.arg state}/{.arg type} for every government matching a predicate.",
+      "i" = "Passing both intersects them: the governments in {.arg govid} that also match {.arg state}/{.arg type}."
+    ), class = "uscogdata_no_cohort")
+  }
+  structure(
+    list(
+      ids        = govid,
+      state      = state,
+      type       = type,
+      state_fips = if (is.null(state)) NULL else .coerce_state_to_fips(state),
+      type_int   = if (is.null(type)) NULL else .coerce_type(type)
+    ),
+    class = "uscogdata_cohort"
+  )
+}
+
+#' Is any part of this cohort expressed as an xwalk predicate?
+#' @noRd
+.cohort_by_predicate <- function(cohort) {
+  !is.null(cohort$state_fips) || !is.null(cohort$type_int)
+}
+
+#' Render the cohort as a SQL boolean expression over `col`.
+#'
+#' `col` may be qualified (`"l.canonical_govid"`, `"x.canonical_govid"`) --
+#' several call sites join the xwalk under an alias. The subquery's own
+#' projected column stays unqualified: it selects from canonical_fips_xwalk,
+#' not from the outer relation.
+#' @noRd
+.cohort_sql <- function(cohort, col = "canonical_govid") {
+  preds <- character(0)
+
+  if (!is.null(cohort$ids)) {
+    preds <- c(preds, sprintf("%s IN (%s)", col, .sql_lit_chr(cohort$ids)))
+  }
+
+  if (.cohort_by_predicate(cohort)) {
+    xwalk_preds <- character(0)
+    if (!is.null(cohort$state_fips)) {
+      xwalk_preds <- c(xwalk_preds,
+                       sprintf("fips_state = %s", .sql_lit_chr(cohort$state_fips)))
+    }
+    if (!is.null(cohort$type_int)) {
+      xwalk_preds <- c(xwalk_preds, sprintf("govs_type = %d", cohort$type_int))
+    }
+    preds <- c(preds, sprintf(
+      "%s IN (SELECT canonical_govid FROM canonical_fips_xwalk WHERE %s)",
+      col, paste(xwalk_preds, collapse = " AND ")
+    ))
+  }
+
+  paste(preds, collapse = " AND ")
+}
+
+#' How many governments the cohort covers.
+#'
+#' One COUNT against the crosswalk, used only to populate the provenance
+#' `scope$cohort` block. Deliberately a count rather than the id list: a
+#' fleet-scale cohort would otherwise put 20,000 ids into every response body,
+#' which is the cost this issue exists to remove.
+#' @noRd
+.cohort_count <- function(con, cohort) {
+  sql <- sprintf(
+    "SELECT COUNT(*) AS n FROM canonical_fips_xwalk WHERE %s",
+    .cohort_sql(cohort)
+  )
+  as.integer(DBI::dbGetQuery(con, sql)$n[[1]])
+}
+
+#' The provenance `scope$cohort` block for a predicate cohort, or NULL when
+#' the cohort was named by id alone (in which case `govids_found`/
+#' `govids_missing` already describe it exactly).
+#' @noRd
+.cohort_provenance <- function(con, cohort) {
+  if (!.cohort_by_predicate(cohort)) return(NULL)
+  list(
+    state         = if (is.null(cohort$state)) NA_character_ else as.character(cohort$state),
+    type          = if (is.null(cohort$type)) NA_character_ else as.character(cohort$type),
+    n_governments = .cohort_count(con, cohort)
+  )
+}
diff --git a/R/complete.R b/R/complete.R
index a5cfc77..5b16b68 100644
--- a/R/complete.R
+++ b/R/complete.R
@@ -57,7 +57,7 @@
 #' aggregate rows. Without it the grid would offer cells the verb structurally
 #' never returns, so every one of them would fill as a phantom $0.
 #' @noRd
-.completion_grid_sql <- function(subtype_col, govid, years, category,
+.completion_grid_sql <- function(subtype_col, cohort, years, category,
                                  subtype_scope) {
   category_pred <- if (is.null(category)) {
     ""
@@ -76,13 +76,13 @@
      JOIN canonical_fips_xwalk x ON x.govs_type = cs.type
      JOIN summary_categories c   ON c.item_code = cs.item_code
      JOIN representation r       ON r.year = cs.year
-     WHERE x.canonical_govid IN (%2$s)
+     WHERE %2$s
        AND cs.year IN (%3$s)
        AND NOT cs.is_aggregate
        AND c.category IS NOT NULL
        AND c.%1$s IN (%4$s)
        %5$s",
-    subtype_col, .sql_lit_chr(govid),
+    subtype_col, .cohort_sql(cohort, "x.canonical_govid"),
     paste(as.integer(years), collapse = ","),
     .sql_lit_chr(subtype_scope), category_pred
   )
@@ -94,10 +94,10 @@
 #' provenance block. Reported rows are passed through untouched -- filling
 #' must never alter or drop what the corpus actually published.
 #' @noRd
-.complete_result <- function(result, con, subtype_col, govid, years, category,
+.complete_result <- function(result, con, subtype_col, cohort, years, category,
                              subtype_scope) {
   grid <- tibble::as_tibble(DBI::dbGetQuery(
-    con, .completion_grid_sql(subtype_col, govid, years, category, subtype_scope)
+    con, .completion_grid_sql(subtype_col, cohort, years, category, subtype_scope)
   ))
 
   result$value_source <- rep("reported", nrow(result))
diff --git a/R/recipes.R b/R/recipes.R
index 11a55be..15a1c6d 100644
--- a/R/recipes.R
+++ b/R/recipes.R
@@ -113,7 +113,7 @@ cog_recipes <- function(pattern = NULL) {
 #' `year_min`/`year_max` -- so there is no double-counting. (Checkpoint
 #' review docs/phase_r_harmonization_review.md § 0.2.)
 #' @noRd
-.run_recipe <- function(con, recipe_id, govid, years) {
+.run_recipe <- function(con, recipe_id, cohort, years) {
   sql <- sprintf(
     "SELECT l.year, l.canonical_govid,
             COALESCE(x.gov_name, l.gov_name) AS gov_name,
@@ -128,11 +128,11 @@ cog_recipes <- function(pattern = NULL) {
            OR (r.gov_type_scope = 'local' AND l.type BETWEEN 1 AND 3))
      LEFT JOIN canonical_fips_xwalk x USING (canonical_govid)
      WHERE r.recipe_id = %1$s
-       AND l.canonical_govid IN (%2$s)
+       AND %2$s
        AND l.year IN (%3$s)
      GROUP BY 1, 2, 3
      ORDER BY 1, 2",
-    .sql_lit_chr(recipe_id), .sql_lit_chr(govid),
+    .sql_lit_chr(recipe_id), .cohort_sql(cohort, "l.canonical_govid"),
     paste(as.integer(years), collapse = ",")
   )
   result <- tibble::as_tibble(DBI::dbGetQuery(con, sql))
diff --git a/R/revenue.R b/R/revenue.R
index 376f9a6..b6efe03 100644
--- a/R/revenue.R
+++ b/R/revenue.R
@@ -50,11 +50,12 @@
 #'   optional `pop_source`, `codes_included`, `aggregate_fallback`, `notes`,
 #'   and `value_source` when `complete = TRUE`.
 #' @export
-cog_revenue <- function(govid, years, category = NULL,
+cog_revenue <- function(govid = NULL, years, category = NULL,
                         per_capita = FALSE, adjust_to_year = NULL,
                         basis = c("harmonized", "raw"), recipe = NULL,
                         revenue_concept = c("general", "total"),
-                        complete = FALSE, limit = NULL, offset = NULL) {
+                        complete = FALSE, limit = NULL, offset = NULL,
+                        state = NULL, type = NULL) {
   # flow_prefixes no longer classifies rows (crosswalk revenue_subtype
   # membership does -- General Revenue, i.e. everything except
   # insurance_trust) -- it only scopes the recipe-suggestion machinery to
@@ -75,6 +76,8 @@ cog_revenue <- function(govid, years, category = NULL,
     revenue_concept = revenue_concept,
     complete       = complete,
     limit          = limit,
-    offset         = offset
+    offset         = offset,
+    state          = state,
+    type           = type
   )
 }
diff --git a/R/search.R b/R/search.R
index 01c27bd..cb3abbd 100644
--- a/R/search.R
+++ b/R/search.R
@@ -184,11 +184,18 @@ cog_gov_search <- function(name = NULL, state = NULL, type = NULL) {
   if (!is.character(state) || length(state) != 1L) {
     cli::cli_abort("`state` must be a 2-letter USPS abbrev or a FIPS integer.")
   }
-  fips <- .state_abbrev_to_fips[[toupper(state)]]
-  if (is.null(fips)) {
-    cli::cli_abort("Unknown state abbreviation: {state}.")
+  # Membership tested before the lookup, not after: `.state_abbrev_to_fips` is
+  # a named CHARACTER vector, and `[[` on a name it does not carry throws
+  # base R's "subscript out of bounds" rather than returning NULL -- which
+  # made the curated message below unreachable dead code. Reported as a bare
+  # subscript error, `cog_gov_search(state = "ZZ")` gave no hint that the
+  # argument wants a postal abbreviation.
+  key <- toupper(state)
+  if (!key %in% names(.state_abbrev_to_fips)) {
+    cli::cli_abort("Unknown state abbreviation: {state}.",
+                   class = "uscogdata_unknown_state")
   }
-  fips
+  .state_abbrev_to_fips[[key]]
 }
 
 # USPS state / territory abbreviation -> 2-digit FIPS code.
diff --git a/R/spending.R b/R/spending.R
index 61a6559..40407a2 100644
--- a/R/spending.R
+++ b/R/spending.R
@@ -68,7 +68,9 @@
 #' millions/billions). The conversion is recorded in the provenance attribute
 #' under `transformations$units_conversion`.
 #'
-#' @param govid Character vector of `canonical_govid` values.
+#' @param govid Character vector of `canonical_govid` values, or `NULL` to name
+#'   the cohort by `state`/`type` instead. One of `govid`, `state`, or `type`
+#'   is required.
 #' @param years Integer vector of years.
 #' @param category Character vector of category names (from
 #'   `summary_categories.category`), or `NULL` for all categories broken out
@@ -185,6 +187,29 @@
 #'   `recipe` and with `complete = TRUE` -- see `offset` and `total_rows`.
 #' @param offset Rows to skip before `limit` starts counting (0-based).
 #'   Ignored if `limit` is `NULL`; defaults to `0L` when `limit` is set.
+#' @param state,type Name the cohort by predicate instead of by id: `state` is
+#'   a 2-letter USPS abbreviation (or a FIPS code) and `type` is one of
+#'   `"state"`, `"county"`, `"city"`, `"township"` (or the integer `0:3`) --
+#'   the same vocabulary, and the same internal coercion, as
+#'   [cog_gov_search()]. Both default to `NULL`.
+#'
+#'   The cohort is then expressed as a subquery against `canonical_fips_xwalk`
+#'   inside each statement rather than round-tripped through R as a literal id
+#'   list. For a fleet-scale cohort that is the difference between a
+#'   301,591-character `IN` list re-parsed in 5--8 statements per call and a
+#'   constant-size predicate: measured at **94 ms versus 449 ms** for the same
+#'   FY2022 aggregate over the 20,106-government `type = "city"` cohort, within
+#'   7% of the no-filter floor.
+#'
+#'   Supplying `govid` **and** `state`/`type` INTERSECTS them -- the
+#'   governments in `govid` that also match the predicate -- rather than one
+#'   silently taking precedence. Naming no cohort at all (`govid`, `state` and
+#'   `type` all `NULL`) aborts with class `uscogdata_no_cohort`.
+#'
+#'   When the cohort is named by predicate, `provenance$scope$govids_found`
+#'   and `govids_missing` are empty -- there is no id list to report against --
+#'   and `provenance$scope$cohort` carries `state`, `type` and
+#'   `n_governments` instead. A `govid`-named cohort reports exactly as before.
 #' @return Tibble with columns `year`, `canonical_govid`, `gov_name`,
 #'   `spend_subtype`, `category`, `amt_nominal`, optional `amt_real`,
 #'   optional `amt_per_capita_nominal`, optional `amt_per_capita_real`,
@@ -198,11 +223,12 @@
 #'   round trip -- so a caller walking pages never has to ask "how many are
 #'   there" separately.
 #' @export
-cog_spending <- function(govid, years, category = NULL,
+cog_spending <- function(govid = NULL, years, category = NULL,
                          per_capita = FALSE, adjust_to_year = NULL,
                          basis = c("harmonized", "raw"), recipe = NULL,
                          expenditure_concept = c("primary", "direct", "total"),
-                         complete = FALSE, limit = NULL, offset = NULL) {
+                         complete = FALSE, limit = NULL, offset = NULL,
+                         state = NULL, type = NULL) {
   # flow_prefixes no longer classifies rows (crosswalk subtype membership
   # does, per expenditure_concept) -- it only scopes the recipe-suggestion
   # machinery to this verb's recipe families (see R/suggestions.R; the
@@ -223,7 +249,9 @@ cog_spending <- function(govid, years, category = NULL,
     expenditure_concept = expenditure_concept,
     complete      = complete,
     limit         = limit,
-    offset        = offset
+    offset        = offset,
+    state         = state,
+    type          = type
   )
 }
 
@@ -251,7 +279,8 @@ cog_spending <- function(govid, years, category = NULL,
                            basis = c("harmonized", "raw"), recipe = NULL,
                            expenditure_concept = c("primary", "direct", "total"),
                            revenue_concept = c("general", "total"),
-                           complete = FALSE, limit = NULL, offset = NULL) {
+                           complete = FALSE, limit = NULL, offset = NULL,
+                           state = NULL, type = NULL) {
   basis_explicit <- length(basis) == 1L
   basis <- match.arg(basis, c("harmonized", "raw"))
   # match.arg() itself throws a base `simpleError`, not an rlang-classed
@@ -291,7 +320,10 @@ cog_spending <- function(govid, years, category = NULL,
     .revenue_concept_subtypes(revenue_concept)
   }
 
-  govid <- .coerce_govid_input(govid, arg = "govid")
+  # NULL `govid` means "the cohort is named by predicate"; anything else is
+  # coerced and validated exactly as before, so an empty or wrong-typed vector
+  # still fails with its original message rather than being read as absent.
+  govid <- if (is.null(govid)) NULL else .coerce_govid_input(govid, arg = "govid")
   # allow_all_categories = TRUE: cog_spending()/cog_revenue() are the two
   # verbs the reserved pseudo-category is defined for. cog_balances() shares
   # this validator but leaves the argument at its FALSE default, so it
@@ -300,6 +332,10 @@ cog_spending <- function(govid, years, category = NULL,
   .validate_verb_inputs(govid, years, category, per_capita, adjust_to_year,
                         recipe, allow_all_categories = TRUE)
 
+  # Built after validation so the argument-shape errors above keep firing
+  # first, and before .ensure_session() so a bad state/type costs no I/O.
+  cohort <- .make_cohort(govid, state, type)
+
   # Recognize the reserved pseudo-category. Detected after type validation so a
   # non-character `category` still fails with the ordinary type error.
   all_categories <- !is.null(category) && .ALL_CATEGORIES %in% category
@@ -407,7 +443,7 @@ cog_spending <- function(govid, years, category = NULL,
     .validate_recipe_id(con, recipe)
     comps <- .recipe_components(con, recipe)
     recipe_label <- comps$label[[1]]
-    result <- .run_recipe(con, recipe, govid, years)
+    result <- .run_recipe(con, recipe, cohort, years)
     sql <- attr(result, "sql_query")
     result <- .shape_recipe_result(result, subtype_col, recipe_label)
     recipe_block <- list(
@@ -423,7 +459,7 @@ cog_spending <- function(govid, years, category = NULL,
     } else {
       NULL
     }
-    sql <- .build_verb_sql(view, subtype_col, govid, years,
+    sql <- .build_verb_sql(view, subtype_col, cohort, years,
                            if (all_categories) NULL else category,
                            ig_view, subtype_scope,
                            all_categories = all_categories,
@@ -441,7 +477,7 @@ cog_spending <- function(govid, years, category = NULL,
       } else {
         count_sql <- sprintf(
           "SELECT COUNT(*) AS n FROM (%s) AS _uncounted",
-          .build_verb_sql(view, subtype_col, govid, years,
+          .build_verb_sql(view, subtype_col, cohort, years,
                           if (all_categories) NULL else category,
                           ig_view, subtype_scope,
                           all_categories = all_categories)
@@ -457,13 +493,13 @@ cog_spending <- function(govid, years, category = NULL,
   # spurious 0.
   completion <- list(applied = FALSE, rows_filled = 0L, absence_means = list())
   if (complete) {
-    result <- .complete_result(result, con, subtype_col, govid, years,
+    result <- .complete_result(result, con, subtype_col, cohort, years,
                                category, subtype_scope)
     completion <- attr(result, ".completion")
     attr(result, ".completion") <- NULL
   }
 
-  if (per_capita) result <- .attach_per_capita(result, con, govid)
+  if (per_capita) result <- .attach_per_capita(result, con)
   if (!is.null(adjust_to_year)) {
     result <- .attach_real_dollars(result, adjust_to_year, per_capita)
   }
@@ -491,7 +527,7 @@ cog_spending <- function(govid, years, category = NULL,
     basis_for_prov <- resolved$basis
     basis_note_for_prov <- resolved$note
     harmonization <- .build_harmonization_block(
-      con, govid, years, resolved, subtype_col, subtype_scope
+      con, cohort, years, resolved, subtype_col, subtype_scope
     )
     # C1(a): gap detection must run against the Direct leg alone. `result`
     # can also carry UNION'd intergovernmental rows (expenditure_concept =
@@ -506,7 +542,7 @@ cog_spending <- function(govid, years, category = NULL,
     } else {
       result
     }
-    suggestions <- .build_suggestions(con, govid, years, category,
+    suggestions <- .build_suggestions(con, cohort, years, category,
                                        direct_leg_result,
                                        resolved$basis, flow_prefixes,
                                        .select_long_view(view_base, resolved$basis),
@@ -611,6 +647,12 @@ cog_spending <- function(govid, years, category = NULL,
   )
   prov$scope$govids_found   <- scope$found
   prov$scope$govids_missing <- scope$missing
+  # A predicate-named cohort has no id list to report found/missing against
+  # (both stay empty), so it describes itself instead. Deliberately a COUNT
+  # rather than the resolved ids: enumerating them would put 20,000 govids in
+  # every fleet-scale response body, which is the cost this path exists to
+  # remove. NULL for a govid-named cohort, so that output is untouched.
+  prov$scope$cohort <- .cohort_provenance(con, cohort)
   attr(result, "provenance") <- prov
   attr(result, ".popyear_range") <- NULL
   # Attached here, after every downstream transform (per_capita/real-dollar
@@ -639,7 +681,11 @@ cog_spending <- function(govid, years, category = NULL,
 .validate_verb_inputs <- function(govid, years, category,
                                   per_capita, adjust_to_year, recipe = NULL,
                                   allow_all_categories = FALSE) {
-  if (!is.character(govid) || length(govid) == 0L) {
+  # NULL is allowed only because the caller has already established that the
+  # cohort is named some other way (`state`/`type`); .make_cohort() is what
+  # refuses a call that names no cohort at all. A supplied-but-empty `govid`
+  # still fails here, exactly as before.
+  if (!is.null(govid) && (!is.character(govid) || length(govid) == 0L)) {
     cli::cli_abort("`govid` must be a non-empty character vector.")
   }
   if (!(is.integer(years) || is.numeric(years)) || length(years) == 0L) {
@@ -740,10 +786,10 @@ cog_spending <- function(govid, years, category = NULL,
 }
 
 #' @noRd
-.build_verb_sql <- function(view, subtype_col, govid, years, category,
+.build_verb_sql <- function(view, subtype_col, cohort, years, category,
                             ig_view = NULL, subtype_scope = NULL,
                             all_categories = FALSE, limit = NULL, offset = NULL) {
-  govid_lit <- .sql_lit_chr(govid)
+  cohort_pred <- .cohort_sql(cohort)
   years_lit <- paste(as.integer(years), collapse = ",")
   # In all-categories mode there is no category filter: the sum is defined by
   # the concept's SUBTYPE allowlist (subtype_pred below), which is the real
@@ -813,13 +859,13 @@ cog_spending <- function(govid, years, category = NULL,
        string_agg(DISTINCT item_code, ',' ORDER BY item_code) AS codes_included,
        bool_or(is_aggregate) AS aggregate_fallback
      FROM %2$s
-     WHERE canonical_govid IN (%3$s)
+     WHERE %3$s
        AND year IN (%4$s)
        %5$s
        %6$s
      GROUP BY year, canonical_govid, gov_name, xwalk_gov_name, %1$s%8$s
      ORDER BY year, canonical_govid, %1$s%8$s",
-    subtype_col, source_expr, govid_lit, years_lit, category_pred, subtype_pred,
+    subtype_col, source_expr, cohort_pred, years_lit, category_pred, subtype_pred,
     category_select, category_group
   )
 
@@ -845,8 +891,16 @@ cog_spending <- function(govid, years, category = NULL,
   }
 }
 
+#' Join population onto a result and derive the per-capita columns.
+#'
+#' The population lookup is keyed on the govids PRESENT IN `result`, not on the
+#' cohort that produced it. Those are the only ones the LEFT JOIN below can
+#' match, so the joined output is identical either way -- but it means this
+#' works unchanged for a cohort named by predicate (where no id list exists in
+#' R at all), and on a paginated call it looks up one page's governments
+#' instead of the whole fleet's.
 #' @noRd
-.attach_per_capita <- function(result, con, govid) {
+.attach_per_capita <- function(result, con) {
   if (nrow(result) == 0L) {
     result$amt_per_capita_nominal <- numeric(0)
     result$pop_source <- character(0)
@@ -859,7 +913,7 @@ cog_spending <- function(govid, years, category = NULL,
      FROM gov_population_yearly
      WHERE canonical_govid IN (%s)
        AND year IN (%s)",
-    .sql_lit_chr(govid), years_lit
+    .sql_lit_chr(unique(result$canonical_govid)), years_lit
   )
   pops <- tibble::as_tibble(DBI::dbGetQuery(con, sql))
   result <- dplyr::left_join(result, pops,
diff --git a/R/suggestions.R b/R/suggestions.R
index b2b0ce4..f089040 100644
--- a/R/suggestions.R
+++ b/R/suggestions.R
@@ -42,8 +42,8 @@
 #' verb call: recipes whose generic join would fill a real gap in `result`.
 #'
 #' @param con Active DuckDB connection.
-#' @param govid Character vector of canonical_govid values (the verb's raw
-#'   `govid`).
+#' @param cohort The verb's cohort object (see `.make_cohort()`), naming the
+#'   governments by id, by state/type predicate, or both.
 #' @param years Integer vector of requested years.
 #' @param category `category` argument as passed to the verb (character
 #'   vector or `NULL`; suggestions are only computed when non-NULL).
@@ -85,7 +85,7 @@
 #'   ig_recipe_id, trigger, suppressed_amount, suppressed_years,
 #'   suppressed_codes)`, possibly empty.
 #' @noRd
-.build_suggestions <- function(con, govid, years, category, result, basis,
+.build_suggestions <- function(con, cohort, years, category, result, basis,
                                 flow_prefixes, long_view,
                                 all_categories = FALSE,
                                 subtype_col = NULL, subtype_scope = NULL) {
@@ -160,7 +160,7 @@
   # violation would kill signposting, the exact failure class uscogdata#9
   # exists to prevent. Owner's call: keep this simple; a batch-aware
   # optimization, if one is worth building, is a separate issue.
-  supp <- .suppressed_components(con, candidates, govid, years, long_view, flow_prefixes)
+  supp <- .suppressed_components(con, candidates, cohort, years, long_view, flow_prefixes)
 
   if (length(gap_years) == 0L && nrow(supp) == 0L) return(list())
 
@@ -188,9 +188,9 @@
              OR (r.gov_type_scope = 'state' AND l.type = 0)
              OR (r.gov_type_scope = 'local' AND l.type BETWEEN 1 AND 3))
        WHERE r.recipe_id IN (%s)
-         AND l.canonical_govid IN (%s)
+         AND %s
          AND l.year IN (%s)",
-      .sql_lit_chr(candidates), .sql_lit_chr(govid),
+      .sql_lit_chr(candidates), .cohort_sql(cohort, "l.canonical_govid"),
       paste(gap_years, collapse = ",")
     ))
   }
diff --git a/R/suppression.R b/R/suppression.R
index e9a01b4..c822fc0 100644
--- a/R/suppression.R
+++ b/R/suppression.R
@@ -50,7 +50,9 @@
 #'
 #' @param con Active DuckDB connection.
 #' @param candidates Character vector of recipe ids to measure.
-#' @param govid Character vector of canonical_govid values.
+#' @param cohort The verb's cohort object (see `.make_cohort()`), rendered
+#'   into the govid predicate on both the outer scan and the restated
+#'   NOT EXISTS filter.
 #' @param years Integer vector of requested years.
 #' @param long_view Name of the verb's long view, from `.select_long_view()`.
 #' @param flow_prefixes The calling verb's own flow-type prefixes (see
@@ -60,7 +62,7 @@
 #'   dollars), `suppressed_codes` (comma-joined, sorted). Zero rows when
 #'   nothing is suppressed.
 #' @noRd
-.suppressed_components <- function(con, candidates, govid, years, long_view,
+.suppressed_components <- function(con, candidates, cohort, years, long_view,
                                     flow_prefixes) {
   empty <- tibble::tibble(
     recipe_id = character(0), year = numeric(0),
@@ -93,7 +95,7 @@
            OR (r.gov_type_scope = 'state' AND l.type = 0)
            OR (r.gov_type_scope = 'local' AND l.type BETWEEN 1 AND 3))
      WHERE r.recipe_id IN (%1$s)
-       AND l.canonical_govid IN (%2$s)
+       AND %2$s
        AND l.year IN (%3$s)
        AND l.amt <> 0
        AND LEFT(r.component_code, 1) IN (%5$s)
@@ -103,13 +105,13 @@
            AND v.year = l.year
            AND v.item_code = l.item_code
            AND v.year IN (%3$s)              -- restated: enables partition pruning (I3a)
-           AND v.canonical_govid IN (%2$s)   -- restated: pushes the govid filter (I3a)
+           AND %6$s   -- restated: pushes the cohort filter (I3a)
        )
      GROUP BY 1, 2
      ORDER BY 1, 2",
-    .sql_lit_chr(candidates), .sql_lit_chr(govid),
+    .sql_lit_chr(candidates), .cohort_sql(cohort, "l.canonical_govid"),
     paste(as.integer(years), collapse = ","), long_view,
-    .sql_lit_chr(flow_prefixes)
+    .sql_lit_chr(flow_prefixes), .cohort_sql(cohort, "v.canonical_govid")
   )
   tibble::as_tibble(DBI::dbGetQuery(con, sql))
 }
diff --git a/man/cog_balances.Rd b/man/cog_balances.Rd
index 4dab4a7..e573552 100644
--- a/man/cog_balances.Rd
+++ b/man/cog_balances.Rd
@@ -5,18 +5,21 @@
 \title{Cash and security holdings for one or more governments}
 \usage{
 cog_balances(
-  govid,
+  govid = NULL,
   years,
   category = NULL,
   per_capita = FALSE,
   adjust_to_year = NULL,
   basis = c("harmonized", "raw"),
-  recipe = NULL
+  recipe = NULL,
+  state = NULL,
+  type = NULL
 )
 }
 \arguments{
 \item{govid}{Canonical govid(s): a character vector, or a data frame with a
-`canonical_govid` column (e.g. from [cog_gov_search()]).}
+`canonical_govid` column (e.g. from [cog_gov_search()]). `NULL` to name
+the cohort by `state`/`type` instead.}
 
 \item{years}{Integer vector of fiscal years.}
 
@@ -49,6 +52,30 @@ and raw space are identical for holdings. Reported in
 \item{recipe}{Optional harmonization recipe id (see [cog_recipes()]).
 `"cash_securities_z77_wide"` and `"cash_securities_z78_wide"` bridge the
 wide era to the modern one.}
+
+\item{state, type}{Name the cohort by predicate instead of by id: `state` is
+  a 2-letter USPS abbreviation (or a FIPS code) and `type` is one of
+  `"state"`, `"county"`, `"city"`, `"township"` (or the integer `0:3`) --
+  the same vocabulary, and the same internal coercion, as
+  [cog_gov_search()]. Both default to `NULL`.
+
+  The cohort is then expressed as a subquery against `canonical_fips_xwalk`
+  inside each statement rather than round-tripped through R as a literal id
+  list. For a fleet-scale cohort that is the difference between a
+  301,591-character `IN` list re-parsed in 5--8 statements per call and a
+  constant-size predicate: measured at **94 ms versus 449 ms** for the same
+  FY2022 aggregate over the 20,106-government `type = "city"` cohort, within
+  7% of the no-filter floor.
+
+  Supplying `govid` **and** `state`/`type` INTERSECTS them -- the
+  governments in `govid` that also match the predicate -- rather than one
+  silently taking precedence. Naming no cohort at all (`govid`, `state` and
+  `type` all `NULL`) aborts with class `uscogdata_no_cohort`.
+
+  When the cohort is named by predicate, `provenance$scope$govids_found`
+  and `govids_missing` are empty -- there is no id list to report against --
+  and `provenance$scope$cohort` carries `state`, `type` and
+  `n_governments` instead. A `govid`-named cohort reports exactly as before.}
 }
 \value{
 Tibble with columns `year`, `canonical_govid`, `gov_name`,
diff --git a/man/cog_revenue.Rd b/man/cog_revenue.Rd
index 940fe20..e94a3d8 100644
--- a/man/cog_revenue.Rd
+++ b/man/cog_revenue.Rd
@@ -5,7 +5,7 @@
 \title{Summarized revenue by category}
 \usage{
 cog_revenue(
-  govid,
+  govid = NULL,
   years,
   category = NULL,
   per_capita = FALSE,
@@ -15,11 +15,15 @@ cog_revenue(
   revenue_concept = c("general", "total"),
   complete = FALSE,
   limit = NULL,
-  offset = NULL
+  offset = NULL,
+  state = NULL,
+  type = NULL
 )
 }
 \arguments{
-\item{govid}{Character vector of `canonical_govid` values.}
+\item{govid}{Character vector of `canonical_govid` values, or `NULL` to name
+the cohort by `state`/`type` instead. One of `govid`, `state`, or `type`
+is required.}
 
 \item{years}{Integer vector of years.}
 
@@ -126,6 +130,30 @@ exactly as before this parameter existed. Mutually exclusive with
 
 \item{offset}{Rows to skip before `limit` starts counting (0-based).
 Ignored if `limit` is `NULL`; defaults to `0L` when `limit` is set.}
+
+\item{state, type}{Name the cohort by predicate instead of by id: `state` is
+  a 2-letter USPS abbreviation (or a FIPS code) and `type` is one of
+  `"state"`, `"county"`, `"city"`, `"township"` (or the integer `0:3`) --
+  the same vocabulary, and the same internal coercion, as
+  [cog_gov_search()]. Both default to `NULL`.
+
+  The cohort is then expressed as a subquery against `canonical_fips_xwalk`
+  inside each statement rather than round-tripped through R as a literal id
+  list. For a fleet-scale cohort that is the difference between a
+  301,591-character `IN` list re-parsed in 5--8 statements per call and a
+  constant-size predicate: measured at **94 ms versus 449 ms** for the same
+  FY2022 aggregate over the 20,106-government `type = "city"` cohort, within
+  7% of the no-filter floor.
+
+  Supplying `govid` **and** `state`/`type` INTERSECTS them -- the
+  governments in `govid` that also match the predicate -- rather than one
+  silently taking precedence. Naming no cohort at all (`govid`, `state` and
+  `type` all `NULL`) aborts with class `uscogdata_no_cohort`.
+
+  When the cohort is named by predicate, `provenance$scope$govids_found`
+  and `govids_missing` are empty -- there is no id list to report against --
+  and `provenance$scope$cohort` carries `state`, `type` and
+  `n_governments` instead. A `govid`-named cohort reports exactly as before.}
 }
 \value{
 Tibble with columns `year`, `canonical_govid`, `gov_name`,
diff --git a/man/cog_spending.Rd b/man/cog_spending.Rd
index f1374e8..9738687 100644
--- a/man/cog_spending.Rd
+++ b/man/cog_spending.Rd
@@ -5,7 +5,7 @@
 \title{Summarized spending by category}
 \usage{
 cog_spending(
-  govid,
+  govid = NULL,
   years,
   category = NULL,
   per_capita = FALSE,
@@ -15,11 +15,15 @@ cog_spending(
   expenditure_concept = c("primary", "direct", "total"),
   complete = FALSE,
   limit = NULL,
-  offset = NULL
+  offset = NULL,
+  state = NULL,
+  type = NULL
 )
 }
 \arguments{
-\item{govid}{Character vector of `canonical_govid` values.}
+\item{govid}{Character vector of `canonical_govid` values, or `NULL` to name
+the cohort by `state`/`type` instead. One of `govid`, `state`, or `type`
+is required.}
 
 \item{years}{Integer vector of years.}
 
@@ -146,6 +150,30 @@ exactly as before this parameter existed. Mutually exclusive with
 
 \item{offset}{Rows to skip before `limit` starts counting (0-based).
 Ignored if `limit` is `NULL`; defaults to `0L` when `limit` is set.}
+
+\item{state, type}{Name the cohort by predicate instead of by id: `state` is
+  a 2-letter USPS abbreviation (or a FIPS code) and `type` is one of
+  `"state"`, `"county"`, `"city"`, `"township"` (or the integer `0:3`) --
+  the same vocabulary, and the same internal coercion, as
+  [cog_gov_search()]. Both default to `NULL`.
+
+  The cohort is then expressed as a subquery against `canonical_fips_xwalk`
+  inside each statement rather than round-tripped through R as a literal id
+  list. For a fleet-scale cohort that is the difference between a
+  301,591-character `IN` list re-parsed in 5--8 statements per call and a
+  constant-size predicate: measured at **94 ms versus 449 ms** for the same
+  FY2022 aggregate over the 20,106-government `type = "city"` cohort, within
+  7% of the no-filter floor.
+
+  Supplying `govid` **and** `state`/`type` INTERSECTS them -- the
+  governments in `govid` that also match the predicate -- rather than one
+  silently taking precedence. Naming no cohort at all (`govid`, `state` and
+  `type` all `NULL`) aborts with class `uscogdata_no_cohort`.
+
+  When the cohort is named by predicate, `provenance$scope$govids_found`
+  and `govids_missing` are empty -- there is no id list to report against --
+  and `provenance$scope$cohort` carries `state`, `type` and
+  `n_governments` instead. A `govid`-named cohort reports exactly as before.}
 }
 \value{
 Tibble with columns `year`, `canonical_govid`, `gov_name`,
diff --git a/tests/testthat/test-all-categories.R b/tests/testthat/test-all-categories.R
index a8ceb08..0052692 100644
--- a/tests/testthat/test-all-categories.R
+++ b/tests/testthat/test-all-categories.R
@@ -4,7 +4,7 @@ test_that(".build_verb_sql emits a literal category and no category filter in al
   sql <- uscogdata:::.build_verb_sql(
     view         = "spending_annotated",
     subtype_col  = "spend_subtype",
-    govid        = "552025209777",
+    cohort = uscogdata:::.make_cohort("552025209777"),
     years        = 2019L,
     category     = NULL,
     subtype_scope = c("operations", "capital"),
@@ -24,7 +24,7 @@ test_that(".build_verb_sql emits a literal category and no category filter in al
 test_that(".build_verb_sql is unchanged when all_categories is FALSE", {
   args <- list(
     view = "spending_annotated", subtype_col = "spend_subtype",
-    govid = "552025209777", years = 2019L, category = NULL,
+    cohort = uscogdata:::.make_cohort("552025209777"), years = 2019L, category = NULL,
     subtype_scope = c("operations", "capital")
   )
   old <- do.call(uscogdata:::.build_verb_sql, args)
@@ -244,7 +244,7 @@ test_that('"All Categories" candidate scoping is symmetric with .build_verb_sql(
   con <- uscogdata:::.ensure_session()
 
   none <- uscogdata:::.build_suggestions(
-    con, govid = "010000226085", years = 2011L,
+    con, cohort = uscogdata:::.make_cohort("010000226085"), years = 2011L,
     category = "All Categories", result = NULL, basis = "harmonized",
     flow_prefixes = c("E", "F", "G"),
     long_view = "spending_long_harmonized",
@@ -255,7 +255,7 @@ test_that('"All Categories" candidate scoping is symmetric with .build_verb_sql(
   expect_length(none, 0L)
 
   scoped <- uscogdata:::.build_suggestions(
-    con, govid = "010000226085", years = 2011L,
+    con, cohort = uscogdata:::.make_cohort("010000226085"), years = 2011L,
     category = "All Categories", result = NULL, basis = "harmonized",
     flow_prefixes = c("E", "F", "G"),
     long_view = "spending_long_harmonized",
diff --git a/tests/testthat/test-cohort-verbs.R b/tests/testthat/test-cohort-verbs.R
new file mode 100644
index 0000000..3b6f655
--- /dev/null
+++ b/tests/testthat/test-cohort-verbs.R
@@ -0,0 +1,218 @@
+# The money verbs accepting a cohort by predicate (uscogdata#58).
+#
+# The load-bearing property is EQUIVALENCE: naming the same set of governments
+# by id and by state/type must return the same rows. Everything else here is
+# about the ways that equivalence could silently break -- the postal/FIPS
+# translation, the intersection rule, and provenance no longer having an id
+# list to describe.
+#
+# Fixture cohorts used: RI (fips 44) cities = 8 governments, DE (fips 10)
+# counties = 3. Small on purpose; the size of the win is measured against the
+# production corpus, not here.
+
+# --- equivalence -----------------------------------------------------------
+
+test_that("a predicate cohort returns exactly what the same ids return", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "RI", type = "city")$canonical_govid
+    expect_gt(length(ids), 1L)
+
+    by_id   <- cog_spending(govid = ids, years = 2019)
+    by_pred <- cog_spending(years = 2019, state = "RI", type = "city")
+
+    # Compare the data itself, ignoring the provenance attribute -- which is
+    # SUPPOSED to differ (see the scope tests below).
+    expect_equal(
+      as.data.frame(by_id[order(by_id$canonical_govid, by_id$category), ]),
+      as.data.frame(by_pred[order(by_pred$canonical_govid, by_pred$category), ]),
+      ignore_attr = TRUE
+    )
+    expect_gt(nrow(by_pred), 0L)
+  })
+})
+
+test_that("equivalence holds for cog_revenue()", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "DE", type = "county")$canonical_govid
+    by_id   <- cog_revenue(govid = ids, years = 2019)
+    by_pred <- cog_revenue(years = 2019, state = "DE", type = "county")
+    expect_equal(nrow(by_id), nrow(by_pred))
+    expect_equal(sum(by_id$amt_nominal), sum(by_pred$amt_nominal))
+  })
+})
+
+test_that("equivalence holds for cog_balances()", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "DE", type = "county")$canonical_govid
+    by_id   <- cog_balances(govid = ids, years = 2019)
+    by_pred <- cog_balances(years = 2019, state = "DE", type = "county")
+    expect_equal(nrow(by_id), nrow(by_pred))
+    expect_equal(sum(by_id$amt_nominal), sum(by_pred$amt_nominal))
+  })
+})
+
+test_that("equivalence survives per_capita, adjust_to_year and pagination", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "RI", type = "city")$canonical_govid
+
+    # per_capita now keys its population lookup on the rows in the result
+    # rather than on the requested cohort; these must stay identical.
+    by_id   <- cog_spending(govid = ids, years = 2019, per_capita = TRUE,
+                            adjust_to_year = 2020)
+    by_pred <- cog_spending(years = 2019, state = "RI", type = "city",
+                            per_capita = TRUE, adjust_to_year = 2020)
+    expect_equal(by_id$amt_per_capita_nominal, by_pred$amt_per_capita_nominal)
+    expect_equal(by_id$amt_per_capita_real, by_pred$amt_per_capita_real)
+    expect_equal(by_id$pop_source, by_pred$pop_source)
+
+    paged_id   <- cog_spending(govid = ids, years = 2019, limit = 5, offset = 5)
+    paged_pred <- cog_spending(years = 2019, state = "RI", type = "city",
+                               limit = 5, offset = 5)
+    expect_equal(as.data.frame(paged_id), as.data.frame(paged_pred),
+                 ignore_attr = TRUE)
+    expect_identical(attr(paged_id, "total_rows"), attr(paged_pred, "total_rows"))
+  })
+})
+
+test_that("a state-only predicate spans every type in that state", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "DE", type = NULL)$canonical_govid
+    by_id   <- cog_spending(govid = ids, years = 2019)
+    by_pred <- cog_spending(years = 2019, state = "DE")
+    expect_equal(nrow(by_id), nrow(by_pred))
+  })
+})
+
+# --- the postal/FIPS trap --------------------------------------------------
+
+test_that("a postal abbreviation resolves to rows, not to silence", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    # canonical_fips_xwalk.fips_state holds "44", not "RI". A predicate built
+    # from the raw parameter matches nothing and returns an empty result that
+    # reads as "these governments reported nothing" -- the exact trap cog-api
+    # hit. A zero-row result here is the regression.
+    r <- cog_spending(years = 2019, state = "RI", type = "city")
+    expect_gt(nrow(r), 0L)
+
+    # And the FIPS form is accepted as the same cohort.
+    expect_equal(nrow(cog_spending(years = 2019, state = "44", type = "city")),
+                 nrow(r))
+  })
+})
+
+test_that("an unknown state abbreviation aborts with a message that names the problem", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    # Regression: `.state_abbrev_to_fips` is a named character vector, so `[[`
+    # on an absent name threw base R's "subscript out of bounds" and the
+    # curated message was unreachable.
+    expect_error(cog_spending(years = 2019, state = "ZZ"),
+                 class = "uscogdata_unknown_state")
+    expect_error(cog_spending(years = 2019, state = "ZZ"),
+                 "Unknown state abbreviation")
+  })
+})
+
+# --- naming the cohort -----------------------------------------------------
+
+test_that("naming no cohort at all is refused", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    expect_error(cog_spending(years = 2019), class = "uscogdata_no_cohort")
+    expect_error(cog_revenue(years = 2019), class = "uscogdata_no_cohort")
+    expect_error(cog_balances(years = 2019), class = "uscogdata_no_cohort")
+  })
+})
+
+test_that("an empty govid vector still fails as it always did", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    # Supplied-but-empty is a caller error, not "cohort named some other way".
+    expect_error(cog_spending(character(0), 2019),
+                 "must be a non-empty character vector")
+  })
+})
+
+test_that("govid and state/type together intersect", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    cities   <- cog_gov_search(NULL, state = "RI", type = "city")$canonical_govid
+    counties <- cog_gov_search(NULL, state = "DE", type = "county")$canonical_govid
+
+    # The documented rule: the governments in `govid` that ALSO match the
+    # predicate -- never one silently taking precedence over the other.
+    both <- cog_spending(govid = c(cities, counties), years = 2019,
+                         state = "RI", type = "city")
+    only <- cog_spending(govid = cities, years = 2019)
+    expect_equal(as.data.frame(both), as.data.frame(only), ignore_attr = TRUE)
+
+    # A disjoint intersection is empty, not "whichever one won".
+    none <- cog_spending(govid = counties, years = 2019,
+                         state = "RI", type = "city")
+    expect_equal(nrow(none), 0L)
+  })
+})
+
+# --- provenance ------------------------------------------------------------
+
+test_that("a govid cohort's provenance scope is untouched", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "DE", type = "county")$canonical_govid
+    prov <- attr(cog_spending(govid = ids, years = 2019), "provenance")
+    expect_setequal(prov$scope$govids_found, ids)
+    expect_length(prov$scope$govids_missing, 0L)
+    # No cohort block: govids_found already describes this cohort exactly.
+    expect_null(prov$scope$cohort)
+  })
+})
+
+test_that("a predicate cohort describes itself instead of listing ids", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    prov <- attr(cog_spending(years = 2019, state = "DE", type = "county"),
+                 "provenance")
+    # Deliberately NOT the resolved id list: a fleet-scale cohort would put
+    # 20,000 govids into every response body.
+    expect_length(prov$scope$govids_found, 0L)
+    expect_length(prov$scope$govids_missing, 0L)
+    expect_identical(prov$scope$cohort$state, "DE")
+    expect_identical(prov$scope$cohort$type, "county")
+    expect_identical(prov$scope$cohort$n_governments, 3L)
+  })
+})
+
+test_that("the cohort block counts the intersection, not the predicate alone", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    counties <- cog_gov_search(NULL, state = "DE", type = "county")$canonical_govid
+    prov <- attr(
+      cog_spending(govid = counties[1], years = 2019, state = "DE", type = "county"),
+      "provenance"
+    )
+    expect_identical(prov$scope$cohort$n_governments, 1L)
+  })
+})
+
+# --- the SQL actually changed ----------------------------------------------
+
+test_that("a predicate cohort never renders the ids into the query", {
+  skip_if_no_corpus()
+  with_fixture_corpus({
+    ids <- cog_gov_search(NULL, state = "RI", type = "city")$canonical_govid
+    prov <- attr(cog_spending(years = 2019, state = "RI", type = "city"),
+                 "provenance")
+    sql <- prov$sql_query %||% prov$sql
+    skip_if(is.null(sql), "provenance carries no SQL for this verb")
+    # The whole point: cohort size does not enter the SQL string.
+    for (id in ids) expect_false(grepl(id, sql, fixed = TRUE))
+    expect_match(sql, "SELECT canonical_govid FROM canonical_fips_xwalk",
+                 fixed = TRUE)
+  })
+})
diff --git a/tests/testthat/test-cohort.R b/tests/testthat/test-cohort.R
new file mode 100644
index 0000000..e8c6b0a
--- /dev/null
+++ b/tests/testthat/test-cohort.R
@@ -0,0 +1,83 @@
+# A cohort is how every verb names the set of governments it queries. It can be
+# named by explicit id, by a predicate over canonical_fips_xwalk, or by both
+# (intersection). These tests cover the SQL construction itself -- pure string
+# building, no corpus needed -- because that is where the postal/FIPS trap and
+# the 40k-literal blowup both live.
+
+test_that(".make_cohort() keeps an explicit id vector as ids", {
+  ch <- .make_cohort(govid = c("550000227544", "060000000001"))
+  expect_identical(ch$ids, c("550000227544", "060000000001"))
+  expect_null(ch$state_fips)
+  expect_null(ch$type_int)
+  expect_false(.cohort_by_predicate(ch))
+})
+
+test_that(".make_cohort() translates a postal abbreviation to FIPS", {
+  # The trap this whole issue exists to avoid: canonical_fips_xwalk.fips_state
+  # holds "55", not "WI". A predicate written against the raw parameter matches
+  # nothing and returns an empty result that reads as "reported nothing".
+  ch <- .make_cohort(state = "WI")
+  expect_identical(ch$state_fips, "55")
+  expect_true(.cohort_by_predicate(ch))
+})
+
+test_that(".make_cohort() translates a type label to its integer code", {
+  expect_identical(.make_cohort(type = "city")$type_int, 2L)
+  expect_identical(.make_cohort(type = "state")$type_int, 0L)
+  expect_identical(.make_cohort(type = 1)$type_int, 1L)
+})
+
+test_that(".make_cohort() reuses the search verb's coercers for invalid input", {
+  expect_error(.make_cohort(state = "ZZ"), "Unknown state abbreviation")
+  expect_error(.make_cohort(type = "special_district"), "Unknown type")
+  # Out-of-scope types (4 = special district, 5 = school district) are refused
+  # by the numeric branch, with the v0.1-scope message.
+  expect_error(.make_cohort(type = 4), "type must be 0, 1, 2, or 3")
+})
+
+test_that(".make_cohort() rejects naming no cohort at all", {
+  expect_error(.make_cohort(), class = "uscogdata_no_cohort")
+})
+
+test_that(".cohort_sql() renders an id cohort as a literal IN list", {
+  sql <- .cohort_sql(.make_cohort(govid = c("a", "b")))
+  expect_identical(sql, "canonical_govid IN ('a','b')")
+})
+
+test_that(".cohort_sql() renders a predicate cohort as an xwalk subquery", {
+  # The point of the issue: the cohort never becomes a literal list, so its
+  # size does not enter the SQL string at all.
+  sql <- .cohort_sql(.make_cohort(state = "WI", type = "city"))
+  expect_match(sql, "SELECT canonical_govid FROM canonical_fips_xwalk", fixed = TRUE)
+  expect_match(sql, "fips_state = '55'", fixed = TRUE)
+  expect_match(sql, "govs_type = 2", fixed = TRUE)
+  expect_false(grepl("'WI'", sql, fixed = TRUE))
+})
+
+test_that(".cohort_sql() renders ids and a predicate as an intersection", {
+  sql <- .cohort_sql(.make_cohort(govid = c("a", "b"), type = "city"))
+  expect_match(sql, "canonical_govid IN ('a','b')", fixed = TRUE)
+  expect_match(sql, "AND canonical_govid IN (SELECT", fixed = TRUE)
+})
+
+test_that(".cohort_sql() honours a column alias", {
+  # Several call sites join the xwalk under an alias (`l.`, `x.`, `v.`), so the
+  # predicate has to be able to name the qualified column.
+  sql <- .cohort_sql(.make_cohort(state = "WI"), col = "l.canonical_govid")
+  expect_match(sql, "l.canonical_govid IN (SELECT", fixed = TRUE)
+  # The subquery's own column stays unqualified -- it selects from the xwalk,
+  # not from the outer relation.
+  expect_match(sql, "SELECT canonical_govid FROM", fixed = TRUE)
+})
+
+test_that(".cohort_sql() escapes quotes in ids", {
+  sql <- .cohort_sql(.make_cohort(govid = "o'brien"))
+  expect_match(sql, "'o''brien'", fixed = TRUE)
+})
+
+test_that("a predicate cohort's SQL does not grow with cohort size", {
+  # The regression this guards: 20,106 ids rendered to a 301,591-character
+  # IN list, embedded in 5-8 statements per call.
+  wide <- .cohort_sql(.make_cohort(state = "CA", type = "city"))
+  expect_lt(nchar(wide), 200L)
+})
diff --git a/tests/testthat/test-recipes.R b/tests/testthat/test-recipes.R
index a251119..fdd8d77 100644
--- a/tests/testthat/test-recipes.R
+++ b/tests/testthat/test-recipes.R
@@ -251,7 +251,7 @@ test_that(".suppressed_components measures the E67/E68 dollars Public Welfare dr
   s <- uscogdata:::.suppressed_components(
     con,
     candidates = c("welfare_cash_e67_wide", "welfare_cash_e68_wide"),
-    govid = "061037123085", years = 2011L,
+    cohort = uscogdata:::.make_cohort("061037123085"), years = 2011L,
     long_view = "spending_long_harmonized",
     flow_prefixes = c("E", "F", "G"))
 
@@ -269,7 +269,7 @@ test_that(".suppressed_components finds nothing in a modern year", {
   s <- uscogdata:::.suppressed_components(
     con,
     candidates = c("welfare_cash_e67_wide", "welfare_cash_e68_wide"),
-    govid = "061037123085", years = 2019L,
+    cohort = uscogdata:::.make_cohort("061037123085"), years = 2019L,
     long_view = "spending_long_harmonized",
     flow_prefixes = c("E", "F", "G"))
   expect_equal(nrow(s), 0L)
@@ -280,7 +280,7 @@ test_that(".suppressed_components rejects a long_view outside the allowlist", {
   con <- uscogdata:::.ensure_session()
   expect_error(
     uscogdata:::.suppressed_components(
-      con, candidates = "welfare_cash_e67_wide", govid = "061037123085",
+      con, candidates = "welfare_cash_e67_wide", cohort = uscogdata:::.make_cohort("061037123085"),
       years = 2011L, long_view = "long; DROP TABLE x",
       flow_prefixes = c("E", "F", "G")),
     class = "uscogdata_internal_error")
@@ -298,7 +298,7 @@ test_that(".suppressed_components never measures a component from the other flow
   s <- uscogdata:::.suppressed_components(
     con,
     candidates = c("welfare_cash_e67_wide", "welfare_cash_e68_wide"),
-    govid = "061037123085", years = 2011L,
+    cohort = uscogdata:::.make_cohort("061037123085"), years = 2011L,
     long_view = "revenue_long_harmonized",
     flow_prefixes = c("T", "A", "U", "B", "C", "D"))
   expect_equal(nrow(s), 0L)