Skip to contents

Title roles: BOARD / OFFICER / STAFF / OTHER

title-roles.Rmd

The title.standard vocabulary from titleclassifier has ~250 values — too fine to reason about the composition of a match. classify_title_role() collapses it into four coarse roles used by cross-organization linkage and interlock studies.

The four roles 

classify_title_role(c(
  "BOARD PRESIDENT", "BOARD SECRETARY", "TREASURER",   # governance -> BOARD
  "CHIEF FINANCIAL OFFICER", "EXECUTIVE DIRECTOR",      # management -> OFFICER
  "VICE PRESIDENT",                                     # governance -> BOARD
  "VICE PRESIDENT OF FINANCE",                          # functional -> STAFF
  "DIRECTOR OF MARKETING",                              # functional -> STAFF
  "CHAPLAIN"                                            # unmatched  -> OTHER
))
#> [1] BOARD   BOARD   BOARD   OFFICER OFFICER BOARD   STAFF   STAFF   OTHER  
#> Levels: BOARD OFFICER STAFF OTHER
  • BOARD — the governance body and its elected officers. Per the interlock use case, the governance triad (president / treasurer / secretary) is BOARD, not OFFICER.
  • OFFICER — paid executive leadership: the C-suite, executive director, general manager, general counsel, comptroller.
  • STAFF — other paid employees: functional directors and vice presidents, managers, coordinators, clerks.
  • OTHER — anything unmatched (e.g. CHAPLAIN, HISTORIAN), by design, so the classifier stays safe on the long tail.

The distinction that does the work

A governance title carrying a functional qualifier is a paid post, not a governance seat. Any title containing " OF " is excluded from BOARD:

classify_title_role(c("VICE PRESIDENT", "VICE PRESIDENT OF FINANCE"))
#> [1] BOARD STAFF
#> Levels: BOARD OFFICER STAFF OTHER

By default the result is an ordered factor (levels = TRUE) so roles sort BOARD < OFFICER < STAFF < OTHER; pass levels = FALSE for a plain character vector.

Inspecting the rules

The rule set is exported and inspectable — the classifier is not a black box:

str(title_role_rules(), max.level = 1)
#> List of 3
#>  $ board  : chr "^((VICE|CO|FIRST|SECOND|THIRD|FOURTH|1ST|2ND|3RD|4TH|SR|JR|SENIOR|JUNIOR|ASSISTANT|ASST|ASSOCIATE|ASSOC|DEPUTY|"| __truncated__
#>  $ officer: chr "^C[A-Z]?[EFOAITDMR]O$|CHIEF .*OFFICER|EXECUTIVE DIRECTOR|EXECUTIVE VICE PRESIDENT|GENERAL MANAGER|MANAGING DIRE"| __truncated__
#>  $ staff  : chr " OF |DIRECTOR$|DIRECTOR |MANAGER|COORDINATOR|ADMINISTRATOR|SUPERVISOR|CLERK|EMPLOYEE|^STAFF$|ASSISTANT|WEBMASTE"| __truncated__

Auditing coverage and pinning exceptions

Because unmatched titles fall to OTHER on purpose, you should audit what lands there for your corpus. title_role_coverage() returns one row per distinct title with its frequency and assigned role:

titles <- c(rep("BOARD PRESIDENT", 5), rep("EXECUTIVE DIRECTOR", 3),
            rep("DIRECTOR OF NURSING", 2), "SERGEANT AT ARMS", "CHAPLAIN")
cov <- title_role_coverage(titles)
cov
#>                 title n    role
#> 1     BOARD PRESIDENT 5   BOARD
#> 2  EXECUTIVE DIRECTOR 3 OFFICER
#> 3 DIRECTOR OF NURSING 2   STAFF
#> 4            CHAPLAIN 1   OTHER
#> 5    SERGEANT AT ARMS 1   OTHER

# The tail worth reviewing:
subset(cov, role == "OTHER")
#>              title n  role
#> 4         CHAPLAIN 1 OTHER
#> 5 SERGEANT AT ARMS 1 OTHER

When you find a title in OTHER that belongs elsewhere, pin it with overrides instead of editing rules — the override wins:

classify_title_role(c("SERGEANT AT ARMS", "CHAPLAIN"),
                    overrides = c("SERGEANT AT ARMS" = "BOARD"))
#> [1] BOARD OTHER
#> Levels: BOARD OFFICER STAFF OTHER

Where roles are used

Role assignment drives the edge filters in cross-organization linkage — board_only_edges() keeps board-to-board interlocks, non_board_edges() the rest. See Cross-organization linkage.

See also