Skip to contents

Step 07 - Standardize titles

Source: vignettes/step-07-standardize-titles.Rmd
step-07-standardize-titles.Rmd

After steps 01-06 a title is clean but still one of thousands of forms. standardize_titles() maps each one to a canonical title.standard, one of a few hundred titles that the taxonomy in step 08 describes. It works in three passes:

  1. C-suite fixes (basic_csuite_fixes()): a few titles are rewritten using pay and the officer box before the lookup. A paid CHANCELLOR is the CEO and an unpaid one the board president; a finance title held by a paid, full-time officer is the CFO.
  2. Exact match: the title is looked up in the title-standardization crosswalk, which lists every known variant. title.match is exact.
  3. Pattern match: a title the crosswalk does not list falls back to a learned rule for its longest known head. “MANAGER IT” ends in no known title but starts with MANAGER, which has a rule. title.match is pattern. Titles whose head depends on context (DIRECTOR, PRESIDENT) have no rule, and a title with no rule stays uncoded (title.match is NA).

Before and after

before (TitleTxt6) title.standard title.match strata strata.label
EXECUTIVE DIRECTOR EXECUTIVE DIRECTOR exact
CEO CEO exact
HEAD OF SCHOOL CEO exact
PASTOR PASTOR exact
DIRECTOR BOARD MEMBER exact
DIRECTOR BOARD MEMBER exact
CHAIR OF BOARD BOARD PRESIDENT exact
CHAIR BOARD PRESIDENT exact
CHAIR BOARD PRESIDENT exact
VICE PRESIDENT BOARD VICE PRESIDENT exact
VICE CHAIR BOARD VICE PRESIDENT exact
TREASURER BOARD TREASURER exact
ASSISTANT TREASURER BOARD TREASURER exact subordinate assistant
SENIOR VICE PRESIDENT OF MARKETING VICE PRESIDENT OF MARKETING exact superior senior
EXECUTIVE VICE PRESIDENT EXECUTIVE VICE PRESIDENT exact superior executive
ADVISORY BOARD MEMBER BOARD MEMBER exact duplex advisory
REGIONAL DIRECTOR BOARD MEMBER exact regional regional
FINANCE COMMITTEE MEMBER BOARD MEMBER exact
STAFF PHYSICIAN DOCTOR exact
HEAD FOOTBALL COACH COACH exact
CAMP DIRECTOR DIRECTOR exact
MANAGER IT MANAGER pattern NA NA
HOSPITALIST PHYSICIAN DOCTOR pattern NA NA
SENIOR RABBI NA NA NA NA
CAMPUS ADVOCATE NA NA NA NA

Things to notice:

  • Many variants collapse onto one standard: DIRECTOR and DIR become BOARD MEMBER, CHAIR and CHAIR OF BOARD become BOARD PRESIDENT.
  • A plain DIRECTOR is a board seat. A director of something (CAMP DIRECTOR, DIRECTOR OF DEVELOPMENT) is a staff job.
  • The crosswalk drops rank words such as SENIOR, EXECUTIVE and ASSISTANT from the standard and keeps them in strata and strata.label: SENIOR VICE PRESIDENT OF MARKETING is a VICE PRESIDENT OF MARKETING at the superior / senior stratum.
  • The standard is the title-only answer. A PRESIDENT may be the paid chief executive or the volunteer board chair. Step 07 files it as BOARD PRESIDENT, and step 09 decides from pay, hours and the checkboxes.

Assets

Asset Type Used for Load it with
title.xwalk crosswalk (package data) every known title.variant and its title.standard, strata and strata.label get_title_xwalk() or data(title.xwalk)
title.patterns rules (package data) head rules for titles the crosswalk does not list get_title_patterns() or data(title.patterns)

Their sources are data-raw/crosswalks/title-standardization.csv and title-patterns.csv in the source repository.

xwalk <- get_title_xwalk()
dim(xwalk)
#> [1] 9181    4
length(unique(xwalk$title.standard))
#> [1] 376

# all the variants that map to one standard
head(xwalk$title.variant[xwalk$title.standard == "CHIEF FINANCIAL OFFICER"], 15)
#>  [1] "CFAO"                                "CFO"                                
#>  [3] "CFO AND ASCENSION LIVING"            "CFO AND CHI HEALTH"                 
#>  [5] "CFO AND CIO"                         "CFO AND CONTINUING CARE"            
#>  [7] "CFO AND TREASURER"                   "CFO AT SSM HEALTH"                  
#>  [9] "CFO AVERA HEALTH"                    "CFO BAYCARE HOSP DIV"               
#> [11] "CFO CFO"                             "CFO OF BHS"                         
#> [13] "CFO OF CHI ST LUKE'S HEALTH SYSTEMS" "CFO OF COMMONBOND"                  
#> [15] "CFO PHC"

# the strata
table(xwalk$strata[xwalk$strata != ""])
#> 
#>   committee      duplex   exofficio    regional   regional? subordinate 
#>         278         138           1         192           1         209 
#>    superior 
#>         358

Each pattern rule names a head word or phrase, where it sits (prefix or suffix), the standard it maps to, and how well it did on titles the crosswalk does list (support titles, precision the share it got right). majority rules map to the most common standard among titles with that head; level rules keep the head’s own standard.

pat <- get_title_patterns()
dim(pat)
#> [1] 121   6
head(pat[order(-pat$support), ], 10)
#>     position            head title.standard     rule support precision
#> 110   suffix  REPRESENTATIVE   BOARD MEMBER majority     300     0.963
#> 65    suffix           CHAIR   BOARD MEMBER majority     261     0.881
#> 116   suffix         TRUSTEE   BOARD MEMBER majority     125     0.864
#> 97    suffix          MEMBER   BOARD MEMBER majority     113     0.912
#> 11    prefix           BOARD   BOARD MEMBER majority     107     0.822
#> 73    suffix       COMMITTEE   BOARD MEMBER majority     107     0.869
#> 63    suffix    BOARD MEMBER   BOARD MEMBER majority      99     0.889
#> 74    suffix COMMITTEE CHAIR   BOARD MEMBER majority      70     0.971
#> 50    prefix         TRUSTEE   BOARD MEMBER majority      68     0.868
#> 34    prefix          MEMBER   BOARD MEMBER majority      65     0.862

Try it 

match_title_patterns(c("MANAGER IT", "HOSPITALIST PHYSICIAN", "SENIOR RABBI"))
#> [1] "MANAGER" "DOCTOR"  NA
title_heads(c("FINANCE COMMITTEE MEMBER", "VICE PRESIDENT OF CLINICAL INTEGRATION"),
            xwalk$title.variant)
#>             suffix                     prefix
#> 1 COMMITTEE MEMBER          FINANCE COMMITTEE
#> 2      INTEGRATION VICE PRESIDENT OF CLINICAL

Input: TitleTxt6. Output: TitleTxt7 (after the C-suite fixes), title.standard, title.match, strata and strata.label. Continue with step 08.