On choosing a constitution (at least the part relating to the distribution of income)

Abstract

A constitution is a collection of principles or axioms determining how society should be organized and a description of the ordering of the axioms in terms of their importance and invocation. We report on an exploratory experiment aimed at discovering preferred axioms relating to the distribution of income within society. Unlike most previous experiments, we inquired directly into preferred axioms, rather than indirectly (done by asking subjects to choose between distributions). In addition to learning that the experimental design was, in principle, appropriate, we discovered that preferences expressed in this direct way appear to differ from preferences expressed indirectly. Interestingly, we also get an insight into the order in which people prefer principles to be implemented, thus suggesting something about relative importance.

Acknowledgement

We thank John Bone for helpful comments and discussions.

Notes

¹The experiment was carried out in Italy. The full instructions are available in Italian at http://www-users.york.ac.uk/~jdh1/hey and pasca mark 1/Istruzioni Trattamento 1 and http://www-users.york.ac.uk/~jdh1/hey and pasca mark 1/Istruzioni Trattamento 2, with English versions at http://www-users.york.ac.uk/~jdh1/hey and pasca mark 1/Instructions Treatment 1 and http://www-users.york.ac.uk/~jdh1/hey and pasca mark 1/Instructions Treatment 2. It should be noted that the principles were first expressed succinctly as above, with more detail being given later. In particular, for principles 1 and 2, subjects were told that this principle would be implemented lexicographically; for principles 3 and 4, subjects were told that the standard deviation would be used as the measure of dispersion; for principles 5 and 6, subjects were told that this principle could be implemented several times; for principles 7 and 8, subjects were told that this principle could be implemented several times and they were asked to specify the group size(s); and for principles 9 and 10, the statement of the principles it was spelt out in detail.

²If it had not, one of the distributions would have been chosen at random; subjects were told about this in the Instructions.

³Note that, had we specified an infinite number, the subjects could have worked out, for example, that one of the possible distributions would be completely equal (€15 for all) and another would be completely unequal (€150 for one and €0 for everyone else). Therefore, once again they would be back to choosing between specific distributions rather than deciding general principles.

⁴One example taken from Amiel and Cowell (1999) is a choice between (1, 4, 7, 10, 13) and (1, 5, 6, 10, 13). Is this a test of the Transfer principle, or of Dispersion or indeed of Rawls?

⁵Consider the distributions (2, 3, 10) and (1, 5, 9). Both 1 and 2 prefer the first.

⁶Or until no agreed-to principles remained (although this never happened).

⁷We note that the subjects repeated the experiment thrice; the results below are based on the aggregate results (there were very few differences between the three repetitions and hence very little evidence of learning).