Sweldex data files — https://sweldex.com
Generated 2026-10-31. License: CC BY 4.0 — cite "Sweldex" with a link (https://sweldex.com).

salary-index.csv   every role x city x remote_scope x month (segments with 10+ ads that state pay)
roles/<role>.csv   the same rows, one file per role
board-medians.csv  median advertised pay on each job board
employers.csv      advertised pay by employer (10+ ads with pay in a role x city, over a 12-month window)

Lines starting with "#" are notes. The first line that does not start with "#" is the header row.

Columns (salary-index.csv, roles/*.csv):
  period_month      YYYY-MM
  role_family       role key; ALL = every role
  city              metro; ALL = whole country
  remote_scope      local = Philippine employer, foreign = overseas client hiring remotely, ALL = both
  n                 ads that state pay (duplicates across boards counted once)
  p25/median/p75    advertised monthly pay in PHP. The published figure is the MEDIAN
  mean              for completeness only; pay is right-skewed, so mean runs above the median
  disclosure_rate   share of all ads in the segment that stated a salary
  release           final / preliminary / in_progress / reconstructed (see the notes in the CSV)
  flags             quality flags, semicolon-separated (empty = none)
  top_source_share  share of the segment's ads that came from its largest job board (0-1, 4 decimals)

IMPORTANT — month-over-month changes:
  A row flagged source_mix_shift must NOT be compared with the previous month.
  The mix of job boards behind it changed sharply, so the change in the median reflects
  what we collected, not the market. The website shows no month-over-month figure for these rows.
  Rows with a high top_source_share (source_dominant) also move with that one board; treat changes with care.

Other flags: source_dominant = one board supplies over 60% of the ads; employer_dominant = one employer supplies a large share;
backfill = most ads were seen 45+ days after posting (collected retrospectively, survivor-biased);
bulk_intake = in a backfill month, one week carries a one-off bulk posting on a board.
Methodology: https://sweldex.com/methodology