election_title: "Hillinger Table 4 — three methods, three winners (EV-3 scale)"

scenario_description: |-
  Claude Hillinger's own worked example, run through the engine. In "Voting and
  the Cardinal Aggregation of Judgments" (Munich Discussion Paper 2004-9, §12,
  Table 4) he uses this 30-voter profile to show what he calls the MIRROR
  PATHOLOGY of STV/IRV: where plurality's famous defect is that an unpopular
  candidate may win, IRV's is that the MOST popular candidate may be eliminated
  in the very first round.

  Ana is ranked first by the fewest voters (9) but is never ranked last by
  anyone. She is the Condorcet winner (beats Bruno 20-10, beats Chloe 19-11) and
  she wins the cardinal count. IRV eliminates her first anyway, and her ballots
  transfer to Bruno.

  The scores are Hillinger's EV-3 ballot shifted onto this repo's 0-5 scale:
  his (-1, 0, +1) written as (0, 1, 2). He states explicitly that the choice of
  origin does not affect the outcome (§5), and the matched file
  hillinger_t4_affine.yaml checks that claim by rescaling to (1, 3, 5).
  Score totals here reproduce his table exactly: Ana 39, Bruno 29, Chloe 22.

  What the engine adds to the paper: Hillinger only reports the STV failure, but
  the same 30 ballots split THREE ways. Plurality elects Chloe (11 first
  choices), RCV-IRV elects Bruno, and STAR / Score / Condorcet all elect Ana.
  One electorate, one set of opinions, three different winners.

  Concept page: 07_Concepts/topics/cardinal_utility.md

paradoxes: [condorcet-winner]
voting_method: STAR
num_winners: 1

# Hillinger (2004), Table 4. His orders are abc / bac / cab; his EV-3 rows are
# (2,1,0) / (1,2,0) / (1,0,2). First choices: Ana 9, Bruno 10, Chloe 11 — so IRV
# eliminates Ana, the one candidate nobody ranks last.
ballots: |-
  Count:Ana,Bruno,Chloe
  9:2,1,0    # Ana > Bruno > Chloe
  10:1,2,0   # Bruno > Ana > Chloe
  11:1,0,2   # Chloe > Ana > Bruno

expected_winners:
  - Ana

# file: hillinger_t4_ev3.yaml