The claim that sortition produces a portrait-in-miniature that “stands for” the target population is categorised by Hanna Pitkin (1967) as a form of “descriptive” representation. I prefer the term “statistical representation” as it makes clear that the reference is to the sample as a whole, rather than the individuals that it is comprised of. There is a temptation to think of sortition as just an alternative mechanism for selecting political officers, and that the end result is still “representatives” akin to the (individual) Honourable Members selected by preference election. But the notion of an (individual) “statistical representative” is clearly an oxymoron. An individual selected as part of a aggregatively-representative sample is just a data point, as in a randomised public opinion survey. In a public opinion survey the views of any individual respondent are of no intrinsic interest, the purpose of the survey is to aggregate individual responses as an indication of the prevalence of different viewpoints within the target population. The fact that individual x has a certain view is irrelevant, all that matters is what proportion of the target population shares the same (or broadly similar) views and the same principle would apply to a representative group constituted by sortition. “Statistical representatives” (to describe the component units of a aggregatively-representative body) is an example of the rare group of terms that only exists in plural form. This places serious constraints on the actions of a body selected by sortition, as statistical representativity only applies at the collective (aggregate) level; indeed it is hard to see what representatives can do other than to register their preferences/beliefs via voting (all votes carrying exactly the same weight), as the differences in the “illocutionary force” of the speech acts of individual members of such assemblies will destroy its aggregative representativity.
Illocutionary force is a linguistic term introduced by the philosopher John L. Austin (1975) that refers to the speaker’s intentions in producing a performative utterance. The illocutionary power of the speech act in persuading others will depend on the speaker’s knowledge, motivation, rhetorical skills and perceived social status. One of the reasons that elected representatives frequently have a background in the legal profession is on account of the rhetorical skills associated with courtroom advocacy. Electors will naturally privilege representatives whose performative utterances are persuasive, on account of the need to establish a parliamentary/congressional majority in favour of their chosen political goals. However, individual members of an assembly chosen by lot are private persons, not representatives, and the speech acts of individual persons will vary widely in their persuasive power. Indeed many randomly-selected persons will lack the confidence and/or motivation to say anything at all. In part 1 of this post I explained how the lot might work as an “invisible hand”, as the law of large numbers will ensure that all factors that might be relevant to political decision making will be represented in the microcosm in proportion to their prevalence in the target population. The larger the sample the more fine-grained the representativity, but the upper limit is constrained by Downs’s principle of rational ignorance — if the sample is too large then individual members will not be motivated to pay attention to the debate as their vote will of be little causal value. But the lack of fine-grained representativity of a sample of (say) 300 pales into insignificance compared to the variation introduced by the speech acts of individual members, hence the need to restrict the remit of the assembly to listening to balanced information/advocacy and then voting in secret. This being the case the arguments set out below do not apply to sortition-based assemblies with a full proposing and deliberative mandate as advocated (for example) by Burnheim (2006), Callenbach and Phillips (2008) and Yoram Gat (frequent posts on this blog).
I have outlined the argument against an active role for a representative body selected by sortition, but we still need to make the case for the judgment of a sample selected by lot as a surrogate for the informed judgment of the target population. It requires nothing more than a simple thought experiment:
- A small group, for example a private club, might be called on to make a decision on a particular matter. After discussing the matter and (if necessary) calling on exogenous information and advocacy, deliberations would end and the matter would be put to the vote (generally by a show of hands). Given that it is not a representative body (all members can attend) the strictures on speech acts given above do not apply. The votes would be counted, the ayes or the nays would have it, and a decision would be made that would indicate the informed preferences of the whole group. Given the majoritarian principle (and the assumption of corporate “personhood”) the decision would be binding on the whole group, irrespective of the personal preferences of each individual.
- Scale up the decision process to a polis (a town, city, district, province or a whole nation) and require all members to attend the debate. Notwithstanding this demanding requirement, the rational ignorance principle would suggest that most members would snooze off, play games on their mobile phones, or chat with their neighbours, in the full knowledge that their individual vote would make little difference. But, say, if in addition to voting all members were required to sit an examination (“failure” leading to the death penalty) to test whether they had attended to the debate and checks were made to see if their vote aligned with their interpretation of the information presented, then the vote tally would be a reliable indication of the informed preferences of the whole group (subject to the majoritarian provisos outlined above).
- If, however, the polis had to defend its borders, monitor its power stations, look after its children and water its crops while the deliberation was going on, it might well choose to draw lots and select (say) 10% of its members to perform these essential tasks, thereby excusing them from political jury duty. Those excused by the process could safely assume that the law of large numbers would ensure that their informed preferences would still be represented by the remaining 90% in proportion to their prevalence throughout the whole population. In other words their personal presence or absence would make no difference to the outcome.
- If the polis were at a low level of technological development, it may well be that (say) 50% of members were required to perform the tasks essential for the maintenance of the state and would need to be excused. Although the level of descriptive accuracy would be slightly reduced, there is no good reason to believe that this would make any significant difference to the outcome, assuming the randomisation principle was sound and no exceptions were made (those of a “political” disposition could not claim exclusion from non-jury duties [watering the crops, securing the borders etc]). A large sample would still be approximately 50/50 on a male-female distribution and there is no reason to believe that the factors that affect political decision-making are not distributed in a similar manner. Note that our interest here is the positive principal of numerical proportionality — i.e. if conservatives outnumbered liberals (say) 3 to 1 in the whole polis then we would anticipate that a large sample would reflect this proportionately (this is the assumption of opinion polls on political issues), so it is a case of the invisible hand, not the blind break. In this sense the lot is highly partial — if it did not select 3 conservatives for every 1 liberal then there would be something wrong with the sampling process (or the size of the sample). Although the blind break ensures impartiality as to which conservatives-minded individuals are selected, we would still anticipate them to outnumber their political opponents at the ratio 3:1.
- Given liberal norms (and practical opportunity costs) modern societies (with the possible exception of Australia) might well baulk at the prospect of compulsory political participation and draconian sanctions on all those who failed to pay attention. Given that the law of large numbers would still apply to a smaller sample, modern liberal societies might use sortition to select a sample that was just large enough to be statistically representative (albeit at a coarser-grained level) but small enough so that those citizens selected by lot would be motivated to stay awake during the debates as their vote would make a real difference; indeed they might even be able to place the casting vote (it’s no coincidence that Athenian political juries comprised between 501 and 5001 members). Sanctions for non-participation would need to be replaced by personal incentives, including compensation for loss of earnings, along with a public decorum that might well include the modern equivalent of the Heliastic Oath along with a republican culture that held up political jury service as a privilege — an essential and honourable civic duty.
- If steps 1-4 in this thought experiment are true, then informed decisions of this microcosm must be held to represent the considered judgment of the whole polis; indeed this hypothesis could easily be tested by drawing multiple concurrent samples from the same target population and seeing if they all came to the same verdict. If so, then this would be an indication of representativity. If they didn’t then the size would need to be increased and/or steps taken to ensure greater consistency of discursive input between groups. No doubt this would further diminish the free and equal exchange of reasons by all participants valued by deliberative democrats, but statistical representativity comes at a cost.
But does this equate to revealing the “will of the people”? The above scenario could only claim to demonstrate the informed judgment of the people (the “disposing” function). The popular will (vox populi, vox dei) also presupposes that policy initiation should be in the hands of the people. The above arguments would suggest that this latter “proposing” function is not amenable to sortition and in this sense I am in agreement with Peter Stone. Because proposing and disposing are fused in most systems of preference election, this has led many to believe that the same would be the case with sortition-based systems. Given the ongoing need for legislative proposals, sortition could only ever be part of a mixed constitutional system and the mixture would need to be based on a separation of functions, as opposed to (say) the hybrid proposals of Callenbach and Philips (2008) and O’Leary (2006), where two congressional houses, one appointed by sortition and one by preference election perform parallel functions.
I had intended to cover the problem of consent as well, but that will require a third post.
John L. Austin (1975), How To Do Things With Words (Oxford University Press).
John Burnheim (2006 ), Is Democracy Possible (University of Sydney Press).
Ernest Callenbach and Michael Phillips (2008), A Citizen Legislature (Imprint Academic).
Kevin O’Leary (2006), Saving Democracy (Stanford University Press).
Hanna Pitkin (1967), The Concept of Representation (University of California Press)