On the model dependence of majority preference relations reconstructed from ballot or survey data

The Condorcet criterion requires a majority winner to be elected whenever it exists. In practice, virtually no social choice procedure or survey research study reports complete pairwise comparisons or a complete ranking of the choice alternatives for each voter, as required by standard majority rule. Thus, majority rule can almost never be computed directly from ballots or survey data. We show that it is typically impossible to unambiguously check a given set of ballot or survey data against the Condorcet criterion because any reconstruction of majority preferences is sensitive to the underlying implicit or explicit model of decision making.  2002 Elsevier Science B.V. All rights reserved.