Conceptual Issues

Historically, the ideas of secrecy and publicity have been understood, and refined, in a number of ways. Before getting to the complications, let me state the obvious: both ideas can be stated as a question of who knows what about whom when. In addition, as we shall see, we may ask who knows who knows what and when (meta-publicity). The knowledge can pertain to a verbal or nonverbal act by a person or to a fact about a person. As examples of an act, consider voting for a candidate in an election or stating the intention to vote for that candidate. As an example of a fact, consider the state of being HIV positive.

An act or a fact is secret if only one person knows about it. With regard to acts, that person is always the agent. With regards to fact, the person may be the affected individual or someone else. I was told in 1999 (I do not know whether it is still true) that in South Africa, doctors were not allowed to communicate the diagnosis to the wife of a man who had been found to be HIV positive, perhaps because fewer men might see a doctor if it were known that their wives would be informed. In this case, the fact was known to two persons: the doctor and the patient. In other cases, a medical fact might be known only to the doctor if she diagnoses a patient with an incurable condition and decides not to tell him. In still other cases, the fact might be known only to the affected person. A person might, for instance, try to hide his color blindness from an employer. In what follows, I shall mostly be concerned with acts, not with facts.

An act or a fact is public, minimally, if at least one other agent knows about it, and maximally if all agents in a relevant group know about it. As we shall see, the relevant group can be small or large; hence the maximum maximorum of publicity obtains when all agents in the largest group (or in the union of all groups) know about it. We can weaken these notions of publicity by requiring only that agents have access to the facts, that is, that they can acquire the knowledge at little or no cost or difficulty if they so desire. The degree of difficulty is important: even facts in the public domain may, for all practical purposes, be secret if one does not know where to look for them. The needle in the haystack is not literally at a secret location.

Benjamin Franklin wrote: “Three can keep a secret if two of them are dead.” He meant, presumably, that minimal publicity would inevitably lead to maximum publicity. Obviously, this is sometimes but not always the case. The saying offers an opportunity, however, to reflect on the cognitive structure of secrecy, by means of three examples listed in increasing order of realism.

If A tells a secret to B, instructing her to keep it to herself, and then learns that C also knows it, A can infer that C heard it from B and may break off his ties with B. Knowing this and valuing his ties with A, B may be reluctant to spill the secret. If, however, A tells the secret to B and C, instructing them to keep it to themselves, and later learns that D also knows it, A may not be able to tell whether D heard it from B or from C. In that case, B or C may be less reluctant to spill the secret. They may, however, be held back by the knowledge that if one of them reveals the secret, the other will know his identity and might break off his ties with the informer (and perhaps tell A). If, finally, A shares the secret with three persons, this case does not arise. To exaggerate the claim: three can keep a secret, but not four.

The point of this analysis can be brought out by a less stylized example. Suppose that on a twelve-member jury it takes ten to convict. An accused person bribes or intimidates three members of the jury to vote Not Guilty (for a real case, see Saunders, Young, and Burton 2010, p. 570), but is nevertheless convicted (in the real case he was acquitted). The accused knows that one of A, B, or C broke his promise, but does not know which of them. If A tells him, credibly, that it was B and the accused then takes revenge on B, the latter will not know whether to blame A or C for causing him to be harmed. If it had taken eleven to convict and only two jurors had been bribed, the informer would not have been able to keep his identity hidden from the promise-breaker.

The chapter by Elster and Le Pillouer provides an even more realistic example. As we explain there, in 1789, the French constituante adopted a paradoxical double tactic: allowing visitors into the assembly where they could observe the votes of individual deputies, while banning the publication of individual votes in the official record. (Bentham, by contrast, argued that both publication and the admission of visitors were essential, the latter to ensure the credibility of the former.) In some cases, visitors printed and circulated lists with the names of deputies who had voted against radical measures. It would obviously have been impossible to identify and take punitive measures against those who leaked the names.

As noted, the question of knowledge can be broken down along two dimensions: known to whom and known when? Consider first debates in a deliberating group, such as a jury or an assembly. In theory, debates could be fully secret even to the members of the group if they took the form of anonymous written exchanges. In practice, this case has probably never been realized (but the Internet creates the potential for its occurrence). Yet even if the debates are public within the group, they may be secret as far as the outside world is concerned. The Federal Convention in Philadelphia provides a famous example, although even this supposedly watertight ship suffered some leaks (Farrand Reference Farrand1966, vol. 3, pp. 61–63). Jury deliberations and discussions on central bank committees have the same dual structure. By contrast, the debates at the Constituante were fully public.

Similarly, voting can take one of three forms: secret even within the group, public within the group but secret to the outside world, and fully public. In this case, the first possibility is not merely theoretical, but instantiated in numerous cases. Today, elections to local or national political bodies are, virtually without exception, based on the secret ballot (see Przeworski's chapter). Some elected bodies have also voted by secret ballot, notably the French National assembly between 1798 and 1845 (Pierre Reference Pierre1893, p. 1019) and the Italian parliament until 1988 (see Giannetti's chapter). On juries, practice apparently varies, although data are scarce. In trade unions and political organizations, there have been recurring conflicts between proponents of secret mail ballots and advocates of voting by the raising of hands at a general meting. As I shall explain shortly, however, the first procedure is not really secret in a rigorous sense.

The second case – internal publicity combined with external secrecy – is illustrated by the Federal Convention, except for the use of secret ballot in electing the members to the important Grand Committees (Elster Reference Elster, Novak and Elster2014). Among central bank committees that decide by voting, most do not publish the names of the members of the minority or the size of the majority. For those who do, the timing of publication is important, as we shall see. Even in this dual case, voting members might deliberately leak information to the outside world, to promote their personal interest or their personal conception of the public interest. Such leaks will usually not, however, be credible: secrecy of debates and votes behind closed doors is self-enforcing.

Bentham's idea of (fully) public voting required publication of individual votes. Historically, this practice has been relatively uncommon. Instead, we observe various forms of “semi-public” voting, such as those discussed in the chapter by Elster and Le Pillouer. Their common feature is that the act of casting one's vote, by rising to one's feet, raising one's hand, going to the left or to the right (the British system of divisions) or answering Yes or No in a roll-call vote, is observable by others. Nobody can count on their vote remaining unknown to their fellow members or to the public at large. Yet to varying degrees, each of these procedures offers some hope of anonymity (and of deniability). At the Constituante, deputies clearly felt less exposed to the public eye if they voted by standing or sitting than if they responded to a roll call. The extreme degree of anonymity in public voting was probably reached in the “shout” that was practiced in ancient Sparta (see Plutarch's “Life of Lycurgus”) and in seventeenth-century England (Kishlansky Reference Kishlansky1986). One could open one's mouth without uttering a sound and go undetected, just as many do today when called upon to sing the national anthem.

The time, if any, at which debates and votes are rendered public, to the internal or to the external audience, can also matter. The time can span seconds, or years. In seriatim voting on a committee – for example, by going around the table – the vote of those who come earlier in the sequence is known to whose who come later. The chapters by Urfalino and Ferejohn discuss this procedure from, respectively, an empirical and a theoretical point of view. In assemblies that vote by a show of hands, it is impossible in practice to prevent some members from taking their cue from others (Bentham Reference Bentham1999, p. 107). The external audience may not learn about the votes cast until much later. After the accidental revelation in 1993 that the Open Market Committee of the Federal Reserve Board had been taping the debates as a help in preparing the minutes and that the tapes had been preserved, Congress pressed the Board into publishing the transcripts with a five-year delay (Meade and Stasavage Reference Meade and Stasavage2008). In 2011, the public learned, for instance, that Ben Bernanke and other board members had minimized the subprime risk in 2006, one year before it became manifest.

Some bodies use secret straw polls, followed first by deliberation and then by a decisive public vote. In his chapter, Vermeule reports that it is also the standard procedure in tenure decisions at Michigan Law School. This practice is apparently common in juries. Some juries may also deliberate before the straw poll. However, to my knowledge, there is no hard evidence about these alleged facts about juries. The secrecy surrounding jury deliberations makes it very difficult to know how they operate. This difficulty is quite general. If debates are not recorded or are conducted behind closed doors, they may not leave any material for empirical analysis. As noted, later accounts by the participants may be unreliable or self-serving. Madison's notes from the Federal Convention provide a rare exception.

The facts that there can be several audiences and that votes can be made public at different times create the possibility for mixed secret-public voting. Bentham (Reference Bentham1999, p. 148) makes the perceptive remark that “[i]n secret voting, the secrecy cannot be too profound: in public voting the publicity can never be too great. The most detrimental arrangement would be that of demi-publicity – as if the votes should be known to the assembly, and should remain unknown to the public.” (In juries, however, that arrangement could be acceptable.) Bentham strongly favored another semi-public regime, which would combine ex ante secrecy with ex post publicity. When casting their votes, nobody would know how others are voting; once the votes are cast, all would learn how others voted. Urfalino's chapter explains how this regime is implemented in the FDA advisory committees. Vermeule, in his chapter, offers a systematic analysis of mixed regimes.

The idea of secrecy can be distinguished, at least roughly, from several related ideas: ignorance, privacy, and anonymity.

In jury trials, the judge often tries to keep the jurors ignorant about certain facts about the accused or about the law, even when these are in the public domain and hence not secret in a strict sense that I define later. In the United States, there is currently a bill before Congress that would create an artificial veil of ignorance by requiring representatives to keep their financial assets in a blind trust. One might also imagine a constitutional reform to the effect that changes in the electoral law will take effect only x + 1 years after being adopted, where x is the length of the electoral cycle. In such cases, what matters is to prevent individuals from learning facts that might shape their actions in undesirable ways.

By contrast, the aim of privacy is to prevent specific others from learning facts that might shape their actions. Insurance companies, for instance, may not be allowed to demand HIV tests as a condition for extending insurance. As noted, doctors may not be allowed to reveal an HIV positive status to a spouse. In almost all countries, the income and wealth of a person are known only to the Internal Revenue Service. In Norway, however, they are accessible on the Internet. It is estimated that, as a result, citizens pay about $100 million dollars more in yearly taxes than they would otherwise have done, perhaps because they fear that neighbors might report a discrepancy between reported income and lifestyle to the tax authorities (Slemrod, Thoresen and Bø Reference Slemrod, Thoresen and Bø2013). In 2014, the government made it possible for taxpayers to learn the identity of those who had checked on them (meta-publicity). The reform caused a reduction of 80 percent in searches. It remains to be seen whether some of the excess tax payments will also be reduced.

Anonymity, finally, pertains to the identity of individuals rather than to their actions. Jurors are sometimes subject to anonymity (based on the fear of retaliation by the defendant or his associates if he is found guilty) in addition to secrecy. In voting, both anonymity and secrecy are usually preserved, but they are not inseparable. To promote higher turnout in elections, one might post the names of nonvoters on the Internet, while respecting the secrecy of their vote – a practice that already exists in Argentina. If this practice were to be generalized and institutionalized, one might supplement it by a regime of meta-publicity by allowing both voters and nonvoters to access the names of those who access their voting behavior.

As I noted earlier, the Norwegian regime of allowing free access to income and tax data had substantial behavioral effects, by increasing tax payments. The knowledge that others might learn one's voting behavior can also modify it. In a large-scale field experiment, “substantially higher turnout was observed among those who received mailings promising to publicize their turnout to their household or their neighbors” (Gerber, Green, and Larimer Reference Gerber, Green and Larimer2008). In both cases, the effect of adding meta-publicity to the first-order publicity remains conjectural.

A regime ensures secrecy “in the strict sense” if it is impossible for an agent to communicate credibly to others how she acted, and thus impossible for others to shape her behavior. A regime in which an agent has the choice between entering a closed booth and dropping a ballot (with a party-specific color) into an urn in the full view of everybody is not secret in this sense. If it is known that supporters of party A will sanction those who vote for party B, choosing the first option will not offer secrecy. Barat's chapter offers an example: “those who take ballots can openly be seen as distrusting the smaller councils” that proposed this optional regime. Nor are mail ballots secret in the strict sense, as shown by Buchstein in his chapter, since supporters of a proposal or candidate might seek out voters to influence them. Even voting officials should not be able to monitor the behavior of the voters, a condition that is violated in juries where each juror communicates her vote orally to the foreperson. Barat's chapter discusses a similar “auricular” vote in Geneva. As he shows, the microtechnology of voting, such as having a curtain between the voter and the person to whom he announces his vote, could be decisive. For Imperial Germany, Leemann and Mares (2011) demonstrate the importance of the color, shape, and transparency of ballots as well as the design of urns. Pierre (Reference Pierre1893) offers numerous examples from France.

Strict secrecy removes credibility from promises to vote for a certain candidate, party, or proposal, and hence removes the incentive to solicit such premises. Yet instead of counting on bribees to keep their promises, bribers might count on their self-interest. In the 2000 Taiwan elections, Kuomintang officials used a gambling scheme to provide monetary incentives for voters to vote for the party's presidential candidate. As one journalist explained, “Organizers for the ruling National Party and local gangsters are offering heavily loaded odds to lure votes to Lien Chan, the party's candidate. Although opinion polls indicate that support is evenly divided in the three-way race, they are promising to pay the equivalent of Pounds 10 for every Pounds 1 bet on a win for Mr. Lien. The odds being offered for the other two candidates are just 80p for every Pounds 1 bet” (London Times, March 15, 2000).

As noted earlier, a voter who opts publicly for secrecy (for herself) in a non-strict regime may by that act reveal her voting intentions. A voter who proposes the secret ballot (for everybody) can run the same risk. He might, therefore, welcome the demand if made by others for the purpose of enabling him to vote according to his true preferences. Thus in the Italian constituent assembly of 1946–48 the Communists demanded secret voting so that Catholics would be free to vote against the proposed indissolubility of marriage (see Giannetti's chapter). The self-defeating nature of a demand for secrecy can be avoided, however, if the demand itself is kept secret. In the nineteenth century, the principle of publicity in the French parliament was often undermined by the transformation of the assembly into a “secret committee” at the request of a small number of deputies (ranging from five to twenty). Before 1870, the names of the requesters were not inserted into the record. Similarly, before the abolition of the right to demand a secret vote, the insertion of the names of the requesters into the record was refused as being contrary to the principle of the secret ballot (Pierre Reference Pierre1893, p. 1010).

Causal Issues

Given a regime of publicity or secrecy, we can ask two causal questions: Why was it adopted? What were its effects? The questions are linked, since a regime may have been adopted for certain intended effects that did in fact materialize, but many effects are either not foreseen or, if foreseen, do not enter among the reasons for adopting the regime.

A main reason for adopting secrecy in elections and juries has always been the desire to protect voters and jurors from bribery and intimidation. In his statistical analysis of the adoption of the secret ballot in national elections, Przeworski finds that both the extension of the suffrage and the introduction of the secret ballot seem to have resulted from the elites yielding to revolutionary threats by the lower classes, but to some extent also from the desire to protect opposition voters from intimidation by incumbents. In his chapter, Barat offers a case study of the introduction of the secret ballot in Geneva that identifies the main cause as the desire of the citizens to reduce the influence of the oligarchic “smaller councils.”

This concern goes back to the classical Athenian democracy. Staveley (Reference Staveley1972, p. 96) writes that the voting procedure used in the Athenian dikasteria in the fifth century BC “was clearly considered to provide some degree of secrecy, and is contrasted by Lysias with the method of placing the ballots on open tables, which he says was introduced by the Thirty deliberately in order to render court proceedings more susceptible to influence.” Although the Athenian assembly usually voted by show of hands, an exception was made for “many specific decisions of the Assembly [that] were required by law to be ratified with a quorum of 6000, voting by ballot and not by show of hands” (Hansen Reference Hansen1991, p. 130). Hansen asserts (ibid.) that “the reason for the special voting-rule was doubtless so that the officials could count the vote and ascertain whether the quorum had been reached.” This may well be the case. It is worth noting, however, that the most important of these specific decisions concerned the granting of citizenship and ostracism. Staveley (Reference Staveley1972, p. 93) claims that the “sources make it clear that the chief purpose behind the use of the ballot…was not to facilitate counting but to ensure secrecy,” since “when the fate of an individual citizen was in the balance, secrecy should be observed wherever possible.” He adds that the secrecy was optional, not mandatory.

I have no competence to adjudicate between these opposed claims. They point, however, to an important distinction between technical and political reasons for adopting a voting regime. Even if the frequent regular votes of the Athenian assembly were taken by show of hands for the technical reason that the secret vote would have taken too much time, it is at least possible that the occasional use of the ballot for specific purposes was politically rather than technically motivated. Similarly, an example in Thucydides I.87.2 suggests that when the Spartans on one important occasion used public voting by division instead of the shout, the reasons were political rather than technical (Lendon Reference Lendon, Tylawski and Weiss2001, p. 174). In the chapter by Elster and Le Pillouer, we argue that the two forms of semi-public voting at the Constituante – by standing or sitting and by roll call – were adopted mainly for technical reasons, related to the desire for vote counting to be rapid when possible and accurate when necessary. The reasons why the roll calls were not followed by the publication of numbers and names were, however, political rather than technical. This example also illustrates the distinction between the reasons for adopting a regime and its effects. Once the Règlement of the assembly had established the rule of voting by standing or sitting, with doubts resolved by a roll-call vote, deputies might demand a roll-call vote to expose or deter those who might vote against radical measures. Similarly, an effect of the public voting in the Athenian assembly was the possibility that dissidents might be intimidated. Thucydides (6.24) states that “with the enthusiasm of the majority [for the Sicilian expedition], the few that liked it not, feared to appear unpatriotic by holding up their hands against it.”

In the case of ostracism, a possible reason for secret voting may have been to prevent a politician from knowing who had voted to ostracize him. That reason was explicit in an unsuccessful attempt by Lord Talbot in 1662 to have Parliament declare some of his opponents incapable of public office:

The scheme came to naught, partly, it seems, for the following reason: “Honours went by ballot at Venice: But punishments had never gone so, since the ostracism at Athens, which was the factious practice of a jealous commonwealth, never to be set up as a precedent under a monarchy” (ibid., p. 212). Juries, including the Athenian jury, have of course practiced punishment by a secret vote. They have done so, however, to decide on guilt, not to select the guilty. In that respect, ostracism may well be unique.

In some cases, the adoption of a public regime may be explained by its potential for intimidation. In my chapter with Le Pillouer, we cite the fact that activists in the student movements of the 1960s and 1970s often imposed a public regime to terrorize opponents. Less anecdotally, we also cite the intervention by the constituant Volney in which he argued for the external publicity of the proceedings on the grounds that it would “shame the perfidious or the coward”; we also cite others who used the same argument to favor plenary debates over committee debates. In communist countries, the optional secrecy in national elections was certainly intended to deter voters from using that option and to create a de facto public regime.

The reasons for adopting internal secrecy of voting in political assemblies vary. In France, under Louis Philippe, the practice was defended by the need to ensure the independence of the chambers vis-à-vis the king (Pierre Reference Pierre1893, pp. 1018–19). Tocqueville (Reference Tocqueville and Tocqueville1985, p. 184) wrote, however, that “[o]ne should not be fooled if a political assembly preferred the secret régime by citing the need to avoid the surveillance by the head of the State: it would only be a pretext. The real motive for this behavior would rather be the desire to submit oneself to his influence without exposing oneself too much to public blame.” As a deputy, he knew the system, and detested it. In her chapter, Giannetti traces the use of secret voting in the Italian parliament back to 1848, and explains the political dynamics by which it was readopted in 1949 by the first elected parliament and finally abolished – by a secret vote – in 1988. By contrast, when the French national assembly abolished the secret vote for its proceedings in 1845, it did so by a public vote (Pierre Reference Pierre1893, p. 1019, n. 2).

Whereas external secrecy in an elected assembly prevents voters from holding their deputies accountable, internal secrecy prevents deputies from trading votes with each other, since under this regime promises to reciprocate have no credibility. As Giannetti shows in her chapter, another effect is that deputies cannot be held accountable by party leaders. At the Federal Convention, the internal secrecy in the election of members to the grand committees may have had the effect of skewing the membership on important issues. Since the votes were cast by delegates, not by states, delegations who for some reason sent many representatives (there were no rules) might shape the composition of the committees to favor the outcome they desired (Elster Reference Elster, Novak and Elster2014).

As Urfalino and Costa show in their chapter, the advisory boards to the Food and Drug Administration have adopted the dual system of ex ante secrecy and ex post publicity advocated by Bentham. Compared to the previous regime of sequential public voting, the effect has been to reduce the number of unanimous votes, suggesting that the reform produced the intended result of reducing conformism. By contrast, as they also note, the comparison does not suggest that previous large majorities were the result of herding and conformism. In other cases, however, public voting might turn a minority into a majority. “Suppose a juror fears that her position is unpopular, or appears insensitive or stupid. Before raising her own hand, she will look around to room to see how many other hands are going up. Other like-minded jurors might be employing the same strategy. The result can be zero votes for a particular verdict, despite the fact that several jurors actually support it” (Schwartz Reference Schwartz2006). There is nothing to exclude that a majority of the jurors might support that verdict but fail to express their support. More generally, ex ante publicity of votes can induce the phenomenon of pluralistic ignorance, notably through the mechanism of informational cascades. As Vermeule notes in his chapter, the risk of cascades is reduced if votes are preceded by debates in which the participants reveal to others the data on which their conclusions are based.

It has been noted in many contexts other than those of debates and voting that the demand for publicity motivated by a desire for transparency can have counterproductive effects. If an agency is required to make its internal documents or deliberations available to the public, the result may be that discussions are moved into other arenas (e.g., “a walk in the park”) where they do not leave a paper trail or an electronic trail. The outcome is a loss of efficiency, since the strategy leads to a defective organizational memory, with no gain in transparency. The requirement that companies disclose the salaries of their directors has led to these compensations spiraling upward, because the salaries set benchmarks for other firms, whereas the intention was the very opposite (Elson and Ferrere Reference Elson and Ferrere2013; see also Surowiecki Reference Surowiecki2013). In the present volume, the chapters by Novak and by de Fine Licht and Naurin illustrate other counterproductive effects of publicity. In Novaks's study of the European Council of Ministers, she shows that the demand for the publicity of voting has caused minorities on the Council to go underground rather than bringing them to the knowledge of the public. De Fine Licht and Naurin canvas four mechanisms that may cause transparency to cause a decrease in public legitimacy rather than, as hoped for, to increase it. They also note, however, that in some cases the expected positive effect does in fact occur.

Normative Issues

Today, the normative question of secret versus public voting in national elections seems settled in favor of secrecy. In his chapter, Manin addresses the main contemporary defense of public voting (Brennan end Pettit 1990) and finds it wanting. Historically, the main defenses of public voting may have been motivated by the desire of elites to control the votes of the nonelite voters. Cicero notoriously opposed the secret ballot on such grounds (see Buchstein Reference Buchstein2000, pp. 140–56 for an analysis). As Buchstein shows in his chapter for the present volume, however, John Stuart Mill defended public voting on the grounds that it would promote the public interest. Even if few today accept his argument against secrecy, one might, perhaps in his spirit, defend lifting the veil of anonymity. Exposing abstainers to the Tribunal of Public Opinion does not seem excessively invasive or coercive. As Justice Scalia said in a different context (Doe v. Reed), “running a democracy takes a certain amount of civic courage.”

Conversely, the normative issue of secret versus public voting in elected assemblies seems now to be settled in favor of internal and (mostly) external publicity. While there may be good reasons for secret sessions of an elected assembly (see Ferejohn's chapter), it is hard to think of good reasons for secret voting. Even in a secret session, there should be internal publicity. A near-universal exception arises, however, in the election of presidents of assemblies. The reason, presumably, is to prevent both intrigues ex ante and resentment ex post. The U.S. House of Representatives is a partial exception to that exception. The following comment by the investigative reporter Matthew Doyle illustrates the complexity of that issue in the context of a plan for unseating John Boehner as Speaker:

Foyle also argues, however, that the roll-call vote on the secret ballot would not necessarily have to be made public. As noted, other assemblies have in fact required that a vote on a secret ballot should itself be kept secret. This requirement seems justified, but only if the idea of having a secret ballot can itself be justified.

The requirement of publicity may not be valid under second-best or nonideal conditions. As Vermeule notes in his chapter, Bentham defended the practice of secret voting in the Polish parliament at a time when Poland was under Russian domination. Under certain conditions, it may be more important to prevent an autocrat from punishing representatives than to ensure that the voters can punish them by non-reelection. Even under ideal conditions one might dispense with publicity: in one-shot constituent assemblies where members do not stand for reelection, external publicity cannot serve the function it has in ordinary legislatures (Elster Reference Elster2013, chapter 4).

If secrecy is the norm in elections and publicity is the norm in assemblies, the situation is less clear for juries, multimember courts, administrative agencies, and expert committees. The great variety observed in practice, both throughout history and in contemporary societies, suggests that there are no compelling normative arguments, or, more plausibly, that there are strong arguments on both sides. In juries, both deliberations and voting must be protected from external publicity. Juries do not – indeed, should not – give reasons for their decisions. Voting by jurors can, however, go together either with internal publicity or with full secrecy. In this case, however, these two regimes can be reconciled through the mixed system proposed by Bentham. In other contexts, the three-stage system practiced at Michigan Law School may serve as a model. The arguments for the mixed regime practiced by the Open Market Committee of the Federal Reserve Bank seem more fragile and conjectural. Ben Bernanke does not seem to have suffered greatly by the revelation of his poor judgment in 2006, and probably did not at the time worry much about being proven wrong five years later.

As a final example, I shall cite the argument made by James d'Angelo (http://worldbitcoinnetwork.com/FAQ/cardboardboxdata.html) that the combination of public voting in Congress and huge private contributions to election campaigns has undermined American democracy, by enabling lobbyists to verify that a representative they have funded votes the way they want. While his argument is persuasive, his remedy – secret voting in Congress – seems excessive. A better solution would be to have political campaigns financed by membership dues and public subsidies.