The African Union has fifty-five member states. The African Continental Free Trade Area is a treaty fifty-four of them have signed. A trade bloc negotiating as one voice is, underneath the diplomacy, an aggregation problem: fifty-five rankings of the alternatives going in, one ranking coming out. Two results from social choice theory bound what any procedure can do with that problem, and this chapter states them exactly, with page numbers, and stops there.
1 · What is established
In 1785, Condorcet showed that pairwise simple-majority voting can produce a social preference with no coherent order at all: alternative A beats B by majority, B beats C by majority, and C beats A by majority, so that the group prefers A to B to C to A. Suzumura's account of it, in the introduction he wrote for the Handbook of Social Choice and Welfare, states the consequence plainly:
Suzumura (2001), on the Condorcet paradox
“He discovered the paradox of voting, or the Condorcet paradox, to the effect that the method of pairwise simple majority voting may yield a social preference cycle — a social alternative A defeating another alternative B by a simple majority, B defeating the third alternative C again by a simple majority, and C in its turn defeating A by a simple majority. … One of the logical implications of the Condorcet paradox is that, once a simple majority cycle occurs … there exists no Condorcet winner — a feasible alternative which is undefeated by any other feasible alternative — thereby excluding the possibility of basing social choice on the seemingly democratic method of collective decision-making.”
In 1951, Kenneth Arrow proved something stronger and more general: no aggregation rule at all — not only pairwise majority — can satisfy two conditions that look minimal without collapsing to a dictatorship. Arrow's own book is not held for this chapter; a 2014 proof by Mark Fey is, and it states the theorem in six lines:
Fey (2014), p. 2, exact statement
“Arrow's Theorem If a social preference function satisfies Unanimity and IIA, then some individual is a dictator.”
Fey's model, stated on the same page: a social preference function maps every profile of individual weak orders over a set of at least three alternatives to a single social weak order. Unanimity says that if every individual ranks x above y, so does society. IIA (Independence of Irrelevant Alternatives) says the social ranking of x against y depends only on how individuals rank x against y, not on where either sits relative to some third alternative. A dictator is a single individual whose strict preference between any two alternatives becomes the social preference, regardless of everyone else. The proof is fully general: weak orders (ties allowed), any finite number of individuals, any finite number of alternatives at or above three.
| claim | state |
|---|
| Condorcet paradox: cyclical majority, three alternatives | established, 1785 |
| Arrow's theorem: Unanimity + IIA ⇒ a dictator | established, 1951; quoted here from Fey 2014 |
| Arrow (1951), Social Choice and Individual Values, itself | not held — see §5 |
| Whether the AU or AfCFTA has hit either failure in practice | not examined — see §5 |
2 · Condorcet's own illustration: two failures, kept apart
Condorcet's Essai sur l'application de l'analyse à la probabilité des décisions rendues à la pluralité des voix (1785) is a 197-page scan with no text layer — a Gallica reproduction of the original, OCR'd for this chapter from the pages cited below. Its Discours préliminaire contains two examples that are often folded into one another. They are not the same failure, and Condorcet does not present them as the same failure.
2.1 · The commerce example, pp. lii–liv: a paradox of propositions, not of candidates
Suzumura calls this “Condorcet's first extended illustration of the paradox of voting,” and it is worth being precise about what it illustrates. Three propositions on trade restriction are put to a vote:
A — any restriction placed on commerce is an injustice.
B — only those restrictions placed through general laws can be just.
C — restrictions placed by particular orders can be just.
Condorcet's own numbers, read from p. liv of the Discours: fifteen voters hold the first combination of views, eleven the second, twelve the third. Tallied proposition by proposition, the combination that wins each individual vote is the one “qui paroissoit avoir le moins de voix” — the one that had appeared to have the fewest votes of the three. Voting proposition-by-proposition and voting combination-by-combination disagree, and the proposition-by-proposition winner is the option that, taken as a whole, the fewest voters actually held.
Why this is a different animal from a three-candidate cycle
This is what a later literature calls a doctrinal or discursive paradox: propositions that entail one another are voted on separately, and majority-consistent votes on the parts assemble into a conclusion no one voted for as a whole. It needs only two live combinations to bite, not three candidates in a cycle, and it does not by itself produce an intransitive social ranking — it produces a winner that contradicts the electorate's own premises.
2.2 · The three-candidate cycle, pp. lvii–lviii: the paradox proper
The cyclical-majority result — the one Suzumura's account above describes, and the one this chapter's §1 states — is a separate passage a few pages later. Condorcet states the general condition first:
Condorcet, Discours préliminaire, p. lvii (OCR, French original)
“… de deux quelconques des trois propositions qui la forment, résulte une conclusion contraire à la troisième …”
— of any two of the three propositions that form it, a conclusion follows that contradicts the third.
Then a worked numerical case, on p. lviii, sixty voters over three candidates A, B and C:
Condorcet, Discours préliminaire, p. lviii (OCR, French original)
“Supposons, par exemple, 60 Votans, dont 23 en faveur de A, 19 en faveur de B & 18 en faveur de C …”
— suppose, for example, 60 voters, of whom 23 favour A, 19 favour B, and 18 favour C.
The OCR breaks off there mid-paragraph; the sixty-voter case goes on to assign each bloc's second preference and derive the cyclical pairwise result. What is established without needing the rest of the paragraph is the shape of the example itself — three candidates, three blocs, none a majority alone — which is the textbook Condorcet cycle, distinct from the commerce example's doctrinal paradox forty pages earlier in the same Discours.
Provenance: how the page numbers were checked
The Gallica scan has PDF page numbers, not the Discours's own roman folios. The offset was read directly off a footnote the scan carries on its PDF page 185, where Condorcet writes of Borda's method: “Cet Ouvrage fut imprimé en entier avant que j'eusse connoissance de cette méthode…” — and Suzumura's own footnote 5 cites that exact sentence as “Condorcet (1785, Discours préliminaire, p.clxxix).” PDF 185 is roman clxxix (179); the offset is roman = PDF − 6 across this printing. Applied backward: PDF 58 is p. lii, PDF 60 is p. liv, PDF 63 is p. lvii, PDF 64 is p. lviii. Every page number in this section is that arithmetic, not a guess at where a phrase “probably” sits.
3 · Why this is Book X's own question
This volume's index states its subject in two readings of one letter: X as ten, the book's number, and X as the Africa term in BRIX — a continent that has ratified a continental trade treaty and not yet acquired one voice to negotiate under it. Fifty-five states, one trade area, one negotiating position: that is an aggregation problem in the exact sense §1 defines one. It does not become a different kind of problem because the inputs are national delegations instead of individual voters, or because the outputs are negotiating positions instead of a ranking of candidates. Arrow's theorem is stated over any finite set of individuals and any finite set of alternatives at or above three; fifty-five delegations and any real trade agenda both clear that bar without needing to strain the model at all.
What the theorem forecloses, precisely: no procedure for turning fifty-five rankings into one collective ranking can guarantee Unanimity and Independence of Irrelevant Alternatives at once, for every possible profile of preferences, without one member's preference becoming the outcome regardless of the other fifty-four. What it does not foreclose: that any particular fifty-five actual delegations, on any particular actual agenda, will ever produce a profile that exposes the failure. Arrow's theorem is a statement about every possible profile; most profiles that occur in practice are far from adversarial enough to trigger it, which is exactly why voting bodies function most of the time. The theorem says the guarantee cannot be built in, not that catastrophe is imminent.
4 · What this chapter does not claim
- That Arrow (1951) has been read. It has not: the book itself is not held for this chapter. Fey's 2014 proof is used instead because it states the theorem exactly, on one page, and is fully general; nothing here depends on any part of Arrow's own text that Fey's restatement does not carry.
- That the African Union's or AfCFTA's actual decision-making procedure has been examined. It has not. This chapter establishes what the mathematics bounds; it does not source the AU's voting rules, quorum requirements, consensus practice, or any instance of a vote on the continent. That is a different, and still open, piece of work — see §6.
- That any real vote, anywhere, has produced a Condorcet cycle or hit Arrow's impossibility. No such claim is made or implied. The theorems bound what is possible in the worst case across all preference profiles; they say nothing about which profiles occur.
- That Suppes's 2005 history of Arrow's intellectual pre-history bears on this chapter's argument. It does not, and is not cited above for that reason: Suppes traces the nineteenth-century development of ordinal utility and the axiomatic method that shaped Arrow's own thinking, and never mentions Condorcet, voting, or the paradox by name. It is held as a separate, genuinely non-overlapping thread, not folded in here to pad the source list.
- That a preprint claiming to overturn Arrow's theorem changes anything above. One such preprint (Lawrence, “Proving Social Choice Possible”) is on file and deliberately not used: an unrefereed claim to overturn a proved theorem is not a source for the theorem, or against it.
5 · What is open
- Arrow (1951) itself. Reading the original would let this chapter cite Arrow directly rather than through Fey's restatement, and would surface whatever context Arrow's own framing carries that a six-page modern proof compresses away.
- The AU's and AfCFTA's actual procedures. Consensus rules, weighted or unweighted voting, quorum and abstention conventions — none of it is sourced here. Without it, §3's claim stays exactly where this chapter leaves it: that the problem has the right shape for the theorems to bind, not that either institution has been measured against them.
- The rest of the sixty-voter paragraph. The OCR of p. lviii stops mid-sentence; Condorcet's own derivation of the cyclical result from the 23/19/18 split has not been transcribed in full.
- A published escape route. Arrow's theorem has a large literature of domain restrictions and relaxed axioms (single-peaked preferences, Black's theorem chief among them) under which majority rule does behave. None of it is surveyed here; if a restriction plausibly describes how trade positions are actually formed, it would belong in a future chapter, not asserted in this one.
6 · Sources
- M.-J.-A.-N. de Condorcet, Essai sur l'application de l'analyse à la probabilité des décisions rendues à la pluralité des voix, 1785. Gallica / Bibliothèque nationale de France scan, 197pp, no text layer; Discours préliminaire pp. lii–liv and pp. lvii–lviii read by OCR for this chapter.
- M. Fey, A Straightforward Proof of Arrow's Theorem, August 18, 2014. Theorem statement p. 2.
- K. Suzumura, Introduction, Handbook of Social Choice and Welfare, Discussion Paper Series A No. 417, 2001. Paradox description and commerce example p. 2; Borda footnote (fn. 5) p. 24.
- P. Suppes, “The pre-history of Kenneth Arrow's social choice and individual values,” Social Choice and Welfare 25:319–326, 2005. Held as a separate thread — see §4.
- K. J. Arrow, Social Choice and Individual Values, 1951. not held.
- This chapter's own ch09-verify.py, which recomputes the hashes, the page offset, and the quoted passages against the files in Downloads.
- This volume, index, §“X is also the Africa term in BRIX”, for the framing §3 rests on.