IS DEMOCRACY FLAWED? I AM NOT ASKING IF DEMOCRACY AS PRACTISED in the UK or USA is flawed, but if the process of democracy itself is flawed as an institution of governing. Winston Churchill famously suggested that democracy was the worst form of government apart from all the rest. Is that really the best we can say for it? In recent years, largely as a result of the massive and growing body of social choice literature, democratic procedures have been seen as necessarily defective. It has been argued that the institutions of both representative and direct democracy cannot be thought to legitimately reveal the true interests of voters or citizens. One of the leading proponents of such a view is William H. Riker.

Riker was an influential political scientist, founder of the ‘Rochester School’, largely though not exclusively composed of academics trained by him.Footnote ² Riker was fascinated by the question of stability. He thought political science the truly dismal science since it is concerned with disequilibrium, whereas economics is concerned with equilibrium. The Rochester School might be defined by its concentration upon the institutions that form to bring order to the social world, a social world that would otherwise be in chaos.

Riker popularized the importance of Arrow's theorem and social choice theory for the study of elections, electoral systems and the processes of politics more generally. He examined log-rolling (or vote-trading), bringing formal analysis to Congressional (and other legislative) studies, and promoted spatial analysis and simple game theory. His influence in all fields of political science is enormous. Indeed, the entire perestroika movement within the American Political Science Association, challenging the dominance of formal and statistical techniques, might be seen as a backlash against the Rochester School.Footnote ³ One aspect of that backlash has been the argument that the Rochester School, in common with one of the other great public choice schools – the Virginians – is sceptical of, if not completely hostile to, the institutions of democracy.

Gerry Mackie's award-winning book Democracy Defended is a full-blooded assault on the Rochester School's ‘irrationalist credo’, most notably contained in Riker's Liberalism Against Populism.Footnote ⁴ It is a powerful indictment of the manner in which empirical claims, which may have little foundation, are repeated when they happen to fit a prevailing opinion. It is less convincing on why Riker might be wrong theoretically. Had Mackie directed his critique to the reasons for the indefensibility of dictatorship against democracy simply on the basis of Arrow's theorem, even under Riker's particular interpretation, then his book would have been more powerful still, and a better defence of democracy.

Why did Riker think that politics was about disequilibrium? His views derive from results in social choice theory and can be most easily explained by considering the Condorcet cycle. For certain consistent and transitive individual preferences, the group preference is intransitive. Suppose we have three committee members, 1, 2 and 3, whose preferences over three alternatives, are respectively a > b > c, b > c > a, and c > a > b (where ‘>’ means strictly preferred to). In a straight majority vote between each pair, a will beat b, c will defeat a and b will beat c. Each alternative can be beaten by another. Any preference aggregation mechanism allowing this cycle to emerge ensures there is no equilibrium. Any mechanism not allowing it to emerge gives an ‘arbitrary’ result. By ‘arbitrary’ Riker does not mean random, or even unpredictable, but rather that the result cannot reflect the ‘true’ structure of the preferences. Under certain preference profiles different mechanisms give different results even with the same set of individual preferences. We have no obvious intuitions as to which mechanisms are the best. Using results from Arrow and McKelvey-Schofield, Riker thought that this was the general form of politics.Footnote ⁵ Hence all electoral results are arbitrary.

Liberalism Against Populism is largely about the normative implications of these results. Broadly speaking, Riker thought they mean that the populist justification for democracy is wrong, and we need a liberal defence (precisely what populism and liberalism mean here will be considered below).Footnote ⁶ He also thought it meant that sometimes ‘heresthetic politicians’ come on to the scene and transform the direction of state policy and political alliances by cleverly manipulating the multi-dimensionality of issue space. That is, they create coalitions by bringing new issues to the forefront of the political battle, breaking up the old coalitions in the process. The Rochester School accepts Riker's normative argument – though it is not their programme. Their programme is the comparative analysis of institutions and of the differences institutions make to the nature of public policy. The common line is that institutions develop in order to bring stability to what would otherwise be chaos.Footnote ⁷ In fact, some within this tradition have argued that Riker misinterprets some of the ‘chaos’ results,Footnote ⁸ leading him to exaggerate the amount of instability. Either way, however, the emphasis is upon institutional analysis, sometimes punctuated by examining special ‘heresthetic’ politicians.

The main target of Mackie's attack is Riker's normative defence of a liberal interpretation of democracy. Riker makes a distinction between populism and liberalism. The distinction is based on how we can interpret the aggregation of preferences and hence the meaning of electoral outcomes. In fact, however, we can find at least three separate understandings of populism in Liberalism Against Populism. Riker first suggests that populism is the idea that democracy should embody the general will:

This definition suggests that there are objectively right answers and objectively correct ways of finding them.Footnote ¹⁰ If there is a right answer, and individuals consider the evidence objectively (and independently), then the general will may indeed be represented by the majority. Modern defenders of this type of populism call it ‘epistemic’ democracy.Footnote ¹¹ In fact Riker does not really attack this type of populism at all, for he immediately shifts attention away from an epistemic ‘objective’ version of voting to a ‘liberal’ preference-based notion where the electorate votes according to their preferences. We should note in passing that ‘preferences’ in this liberal interpretation do not necessarily entail self-interested preferences. A voter's preference might be their judgement about what is best for society as a whole, rather than what is best for them personally. Nevertheless, it is preferences that determine the way people vote, not judgements about the truth of a proposition. Perhaps Riker cannot envisage politics in its ‘epistemic’ form, and hence switches to a more liberal interpretation of voting when constructing populism.

Within the liberal interpretation of voting Riker distinguishes two types of populism, which I shall call the strong and the weak forms. If the epistemic conception is the objective representation of objective interests, the strong form is the objective representation of the subjective interests (or preferences) of people, with the weak form being ‘a’ representation of those preferences. Riker attacked the strong form by arguing that the populist ideal of respresenting the ‘will’ of the people, now translated into the correct ‘representation’ of the preferences of the people, is impossible. He took this as a direct consequence of Arrow's general possibility theorem, which shows that any way of aggregating votes that satisfies three axioms must be dictatorial. The implication that may be taken from Arrow's theorem is that any non-dictatorial social decision function determines the result of any vote as much as the voters' preferences do. Another way of putting this point is to say that there is no social welfare function that is the unique aggregation of the preferences of the people, and so there is no unique social decision function which can represent them. The populist ideal of the general will, at least under the liberal interpretation of voting based on preferences, is impossible.

However, Riker did not leave the matter there. Once he had shown that populism under the strong interpretation was impossible, he further criticizes the weak version. For he imagined that a populist might recognize the truth of this implication from Arrow, yet suggest that some ways of aggregating preferences are superior to others. So one populist might think that we should use some Condorcetian social decision mechanism, another might prefer a different mechanism such as the Borda count. Either way, the result can be considered the representation of the common will. The two different rules will not always pick the same winner, but each populist can claim that, given the rules we have adopted, we can consider the winner to be the general will.

Riker argued that this weakened version of populism is still impossible in the sense that we cannot think that the results of any given vote necessarily reflect the ‘true’ or ‘sincere’ preferences of the electorate. He said that, since all social decision functions are manipulable, we can never know whether any actual vote truly represents what, under sincere voting, the winner would be. Manipulation here means one or both of two things. First, manipulation may occur through agenda-setting. Second, it may occur through strategic voting. Put these together and we have the ‘irrationalist’ claims about the ‘meaninglessness’ of democracy that Mackie quotes at the opening of his book.Footnote ¹² Democracy is ‘meaningless’ since we cannot claim that the result reflects the ‘will of the people’ (1) because there is no such thing (via Arrow), and (2) because we cannot tell if the result was manipulated. And Riker claims that manipulation is ubiquitous. Gerry Mackie defends the populist interpretation of democracy against both claims. But before we get to Mackie's arguments, we should briefly consider Riker's defence of democracy, under the ‘liberal’ interpretation of voting.

Riker defends a liberal conception of democracy. His liberal conception is no more than that the electorate can ‘throw the rascals out’.Footnote ¹³ Any representative who does badly – however that is thought of by the electorate – can be removed from office. We do not need to think that what the representative does on behalf of his constituents actually ‘represents’ anything resembling their ‘common’ or majority view. However, if the representative is inefficient or corrupt he is more likely to be removed from office. He should realize this, and so this fact acts as a check upon his activity. That is all that is required in liberal democracy. So, in the liberal view, the central point of voting is not to produce a specific outcome representing the will of the people but ‘merely’ to keep a check on rulers, which helps to protect the liberty of subjects. Riker distinguishes two possible dangers to liberty: oppression and inefficiency. The threat of being removed from office prevents infringements on subjects' liberty and will make politicians more efficient.Footnote ¹⁴

One obvious criticism of this liberal defence is that if electoral outcomes are as ‘arbritrary’ as the Rochester School sometimes maintains, then a representative need not worry about trying to keep the public happy, since he might be thrown out no matter what he does.Footnote ¹⁵ But, remember, ‘arbitrary’ does not mean random. It means the result does not necessarily reflect the sincere preferences of the electorate. Riker admits that good (non-oppressive and efficient) politicians may be thrown out in favour of bad ones. But all he needs to argue is that in order to give incentives for politicians to be good there needs to be a positive correlation between how relatively good they are and the probability of their being ejected. After all, if there are shifting coalitions, or cycling majorities, then one needs to target some voters, though not all. It is certainly true that the representative might also decide to please no one at all, and take bribes from the very beginning. The voters may decide to punish such a corrupt politician, although they cannot be sure that the challenger might not also be corrupt. However, we can expect voters to be able to distinguish good from bad politicians, and anticipate that corrupt politicians will try to behave like good ones in order to be re-elected.Footnote ¹⁶ Here, perhaps, the electoral system (in its broadest meaning, including informational channels as well as voting rules) will affect how good politicians are. Some systems may have a stronger correlation between bad politicians and punishment than others.Footnote ¹⁷

Mackie makes much of what appears to be straightforward fallacious reasoning in Riker. Because sometimes we cannot be sure that a result is not manipulated we cannot ever be sure a result is not manipulated. Consider this statement: because I am not sure what my wife just said, I can never be sure what she has just said. Not an inference one is likely to get away with in the long run. At times Riker does make inferences that seem to derive from this fallacious reasoning. However, I will argue that his major claim does not have this character.

Mackie is not satisfied with the liberal defence of democracy. He wants populism, so takes on not just Riker but Arrow too. First let us consider Mackie's defence of the ‘true’ representation of preferences. Recall that Arrow's theorem suggests that no social decision mechanism truly represents the common will because, given the possibility of any profile of preferences, different mechanisms may produce different results.Footnote ¹⁸ Since there is no way of measuring what the ‘correct’ result is without using a social decision mechanism, we cannot test the different mechanisms against one another to see which is the best representation. All we can have are other criteria to choose between them.

Arrow's theorem was developed using four axioms that are often defended as being intuitively reasonable. Universal domain (U) says that any preference profile is allowable. The Pareto prinicple (P) says that if everyone prefers x to y, then any method of group decision must rank x above y. Non-dictatorship (D) says that there should not be one person who decides the result. Each seems intuitively compelling, though we may note in passing that condition U may seem reasonable as a theoretical condition, but ordinarily we might think that there are good reasons why people tend to order preferences in similar ways. If that is generally true, then the results of Arrow's theorem may not be so troubling as Riker claims.Footnote ¹⁹ The Independence of Irrelevant Alternatives (IIA) is less intuitively compelling, and, like Mackie, we shall spend some time on it. Arrow's theorem shows that the other three conditions are only mutually compatible if condition D is broken.

Mackie's tactics are to suggest that there is no warrant for claiming that a reasonable decision mechanism should not break condition IIA. He argues that intuitive defences of the independence condition in fact misunderstand what it means in Arrow's theorem. Mackie then proceeds to argue that even under those misunderstandings there is nothing wrong in breaking independence, and certainly when understood correctly there is nothing wrong in breaking IIA. In fact in the literature there are numerous conditions known as ‘the independence condition’. Mackie considers two of them: IIA(A) where ‘(A)’ stands for ‘Arrow’, and IIA(RM) where ‘(RM)’ stands for ‘Radner and Marshak’.Footnote ²⁰ He argues that the latter, also better known as ‘contraction consistency’, is the version of the independence condition that social choice writers generally defend as being ‘intuitively obvious’. Mackie states that Arrow requires condition IIA(A) to prove his theorem, and he suggests that this condition is not intuitively obvious, if only for the fact that so many clever people once confused it with IIA(RM). He also quotes Barry and Hardin, ‘Nobody has any immediate views about the desirability of, say, the independence of irrelevant alternatives, and we should refuse to be bullied by a priori argument to the effect that we would be “irrational” not to accept it'.Footnote ²¹

Whilst the issue, at some level, is about rationality, I do not think it is about whether or not we find some particular example intuitively compelling. I think contraction consistency is required as a condition of rationality in order to be able to interpret someone's actions.Footnote ²² Predictability is a condition of interpretation.Footnote ²³ Only by assuming that people are rational – that is, predictable in the sense of IIA(RM) – are we in a position to discover what their reasons are when they appear to be breaking the axiom. We need to assume contraction consistency in order to be able to work out what someone is doing. Any apparent violation can only be interpreted by showing in what sense it is not a violation.Footnote ²⁴ And the reason why it is not a violation would normally be that one of the alternatives has taken on a new meaning for the individual. We must discover that meaning, but we can only realize that it has a new meaning (and hence interpret their actions) by assuming that they are (contraction) consistent. Mackie has such an example of why ‘contraction consistency’ is not irrational.

His example (from Arrow) is someone who orders three states of the world: Cold War > Hot War > Disarmament. He says, ‘She least prefers Disarmament as that would amount to surrender to the enemy, but also thinks Cold War is better than Hot War because there are fewer casualties in Cold War’.Footnote ²⁵ We now contract the set of three alternatives to two. She still prefers Cold War to Hot War, when faced with just these two alternatives; however, she prefers Disarmament to Cold War when faced just with these. She seems to break contraction consistency. But Mackie says her choice is perfectly rational. ‘If Hot War were off the menu of choice, if Hot War were no longer possible, then the peace of Disarmament would be preferable to the tension of Cold War and would not require surrender to the enemy’.Footnote ²⁶ Quite. ‘Cold War’ does not mean the same when the menu is contracted. Without the threat of Hot War, disarming does not mean ‘surrender to the enemy’. We rationalize her preferences, we explain them, we make sense of them, by showing they do not break contraction consistency. They do not, since ‘Disarmament’ in the set {Cold War, Disarmament} does not mean the same as ‘Disarmament’ in the set {Hot War, Cold War, Disarmament}; hence the first is not a subset of the second. Without the justification provided by Mackie for her orderings we would be puzzled about her preferences and neither she nor we would have any way of predicting her actions.Footnote ²⁷ Contraction consistency, an axiom of rational choice, forces us to examine apparantly inconsistent preferences in order to explain why they are not inconsistent. The axioms of rational choice allow us to examine actions, and to explain them rationally.Footnote ²⁸

Arrow's theorem does not rely on contraction consistency, however, but on the stronger condition IIA(A). Mackie's swipe at IIA(RM) is a mere aside used to attack the early justifications of Arrow's theorem. Once people realize that Arrow requires IIA(RM), Mackie suggests they should revise their views. But rather what happens is ‘a new attempt to justify the newly understood IIA(A). The conclusion is driving the premises, the tail is wagging the dog.’Footnote ²⁹ IIA(A) is based upon binary comparisons so if alternative x is preferred to y, then x is preferred to y when the ordering is xyz, xzy or zxy. Mackie's attempt at showing that violating IIA(A) is ‘substantially rational’ involves a collective, rather than an individual choice. This is not insignificant. Rational choice requires that individuals follow the axioms of rational choice in order to be predictable.Footnote ³⁰ It also predicts that collectives will not always follow the axioms and hence often be unpredictable.Footnote ³¹ But Mackie uses the example normatively. We are faced with a claimed violation of IIA(A) and then invited to think that the violation is, normatively speaking, OK. In other words, Mackie wants to suggest that it is not intuitively compelling that the winner in a binary comparison of two alternatives should not be reversed when the two alternatives are joined by others. The example is illustrated in Table 1.

Table 1 The Condorcet Winner and the Borda Rule

Imagine a reception with nine people, five from the Business School and four from the Law School. The caterer will only provide one type of beverage, so the organizer arranges a poll of those attending to decide which one to choose. The results of the poll are given in the first two columns of Table 1 (for the moment ignore the last column). Beer is chosen since that is preferred by five (of the Business School) to four (lawyers) for coffee. Now, however, the Law School drops out and the Theology School replaces them. Beer is still preferred by five to four over coffee, so the organizer sticks with beer. The theologians are furious. Beer is their least preferred option, since all are teetotal. Mackie says, ‘The organizer looks only at the relevant alternatives, coffee and beer: by pairwise comparisons nothing has changed, beer is still the choice by majority rule.’Footnote ³³ Beer is the Condorcet winner – it beats all other altenatives in pairwise comparisons – whether we consider the Law or Theology School. But under the Borda rule – which assigns a score to each ranking (0 to the lowest, 1 to the second lowest and so on – see the fourth column in Table 1), beer is the winner when the Business and Law Schools vote, but coffee the winner when Theology replaces Law.Footnote ³⁴ The Borda rule can take account of the fact that the Theology School put beer last. Condorcet cannot.

Here we enter a debate. Should we take into account the intensity of preferences? And if so, how do we measure them? Let us add four other alternatives (wine, gin, whisky and vodka) to get Table 2 (excluding the Law School). Ignore for the moment the third column. The Business School all appreciate the extra alcoholic drinks (though not as much as beer), but the theologians do not (indeed they are all worse than beer). We find that even under Borda beer wins, with 61 points ((5 × 9) + (4 × 4)), while coffee scores 56 ((5 × 4) + (4 × 9)). Should this Borda count result satisfy the theologians? No, they will be just as as angry, since they will still have nothing to drink at the reception. The reason, of course, is that the Borda count does not really measure intensity of preference. It is simply an ordinal measure where distance between alternatives depends as much upon the number of alternatives as any ‘intensity’ of preference we have between them. For the theologians all thealcoholic drinks are on a par with each other. Beer may be slightly more amenable since it is less alcoholic, but unless needs must (there is no clean water and no non-alcoholic drinks on the desert island) they could not contemplate drinking it.

Table 2 Adding Alternatives and Strategic Voting

I take this to be the implication that Riker wanted to draw from Arrow's theorem with regard to the general will. There is no representation of the general will through ordinal measures of preference. We may have measures, but it is not obvious that any one is preferable to the others. If it were obvious, then one could argue that one method of voting is superior. But one cannot. Borda may be superior to other rules for all sorts of reasons, but it is not obviously better, and one reason for that is that it breaks IIA(A).Footnote ³⁵ It may not be intuitively obvious that independence is ‘substantially rational’, but it does seem obvious that there are contradictory intuitions regarding the examples in Tables 1 and 2. Even if you thought that the theologians' wishes should take precedence over those of the Business School, it could not be because the ‘general will’ should triumph and the Borda count correctly represents it. Even adding the alternatives in Table 2 our intuition that the theologians' wishes should be respected is equally strong. Any ordinal characterization will fail to capture such ‘urgent’ preferences. And that is one reason why liberals often think that such preferences should be taken out of ordinary democratic procedures and set aside for special constitutional consideration.Footnote ³⁶ It is not lack of correspondence with the general will that leads to our sympathy with the theologians, it is our view that the fact they will not drink alcohol should be given greater weight than allowed in any system of voting that gives each voter equal weighting. The only way to protect the theologians is to give them special rights because we respect their special wishes with regard to drinking alcohol. At least that is the liberal way.Footnote ³⁷

Mackie's objection cannot simply be that ordinal rankings lose information on intensity of preference. We all know that. He makes the somewhat dubious claim that the Borda rule is ‘more cardinal’ than, say a plurality rule, or the Condorcet method. I am not sure something can be more or less cardinal. Either it is cardinal or it is not. It is true that any preference rule carries more information than any plurality rule, but both the Condorcet and Borda rules are preferences rules. They just carry different information.

Nevertheless, Mackie is certainly right that there are good reasons for preferring one method over another, and he does produce some persuasive arguments for the Borda rule.Footnote ³⁸ But then Riker's second objection kicks in: manipulation. Manipulation, recall, comes in two forms: agenda-setting and strategic voting. In the examples in Tables 1 and 2, the reception organizer or caterer could be the agenda setter. Say it was the turn of the Business and Theology Schools to go to the reception. The organizer might have known that the theologians would not want alcohol, but he wants alcohol served (he has a preference for any kind of alcoholic drink). Vice-Chancellor Mackie has already decreed that democratic votes using the Borda rule must be followed for all university decision-making. So the organizer persuades the caterer to offer the longer list of beverages, with numerous alcoholic drinks, believing that the organized drunks of the Business School will ensure they rank order alcoholic drinks to ensure their favoured one is chosen.

If the theologians realize what the organizer is up to they can respond. They can manipulate by strategic voting, and reverse the order of their preferences for alcoholic drinks as represented in column 3 of Table 2. Under the Borda rule coffee now wins again.Footnote ³⁹ Agenda-setting manipulation can be countered by strategic voting. All systems are manipulable (apart from lotteries, where you vote your preference on a ballot and then the ballot is chosen randomly), but defenders of Condorcet argue that Borda is easier to manipulate than some other systems. And Mackie is troubled by manipulation. A large part of his book (eight out of 18 chapters) is devoted to arguing that manipulation rarely occurs in real politics, so we need not be as concerned about its evil possibilities as most social choice writers proclaim.

The Rochester School claims that manipulation is ubiquitous. Critics claim that the Rochester School, surveying nearly 3,000 years of political activity, has only managed to come up with a dozen or so examples of agenda manipulation. With careful scholarship and a fine eye for detail, Mackie examines all these cases and for each argues that either the alleged manipulation did not take place or the evidence for it is exceedingly weak. This is undoubtedly the finest part of the book, using careful scholarship, detailed study, demonstrating deep knowledge of the literature and written with wit and elegance. Mackie's case is that manipulation through agenda-setting is unknown. He does not deny that it may be attempted. But he argues that it has never been shown to have succeeded. Professional politicians are simply too canny to have the wool pulled over their eyes. He is also sanguine about any future attempts to show examples of agenda manipulation. His argument is not that agenda manipulation could never occur; he might even be prepared to concede that it has. But it is not ubiquitous. This is his strongest argument.

One of Riker's claims is that populism is ‘meaningless’. We cannot tell what people really think from the result of a vote. Voting does not reveal preferences. At one level this argument is ridiculous. Mackie argues that the attempt to prove manipulation occurs through agenda-setting is inconsistent with the claim that we cannot tell what people's preferences are, since manipulation requires the agenda setter to have a good idea of the distribution of preferences. If successful agenda-setting is rare, then, while we should be on the look-out for it, we need not be too anxious about its presence.Footnote ⁴⁰

Even though agenda-setting manipulation might be rare, strategic voting might be ubiquitous. It might be argued that the incidence of tactical voting in elections is quite low – let us guess 5–10 per cent at most. How important it has been over the years is open to question. But here we must return to Arrow's theorem. Surely we all vote tactically nearly, if not every time we vote. Tactical voting is voting for a candidate whom one thinks stands the best chance of winning, over those one most prefers. We virtually always vote for a candidate on the ballot paper. But how often do ballot papers contain the names of the person we would most prefer? Writing in names is allowed on some ballots. It is not on others. Either way, writing a name in is unlikely to get that person elected. So we vote tactically.

Is this argument absurd? When considering the alternatives we must consider only those that are truly alternatives. But Arrow's theorem assumes the condition of universal domain (U) that states that any preference ordering is allowed. We know one way to reduce its impact is to reduce the scope of U. One can always avoid voting paradoxes by restricting the field to just two candidates. Restricting candidates to those who happen to want to stand, or have a party to back them, or the money to run themselves is another way of restricting the alternatives. Of course the incidence of cycles is reduced by restricting candidates to those on the ballot. But is this not Riker's point? Voting does not reveal the general will, but rather allows voters to reject candidates. And that is all we can expect it to do. According to Riker (and not challenged by Mackie) the general will is the sincere preferences of the electorate, and those surely are given by their complete orderings of ‘all social states’ in the nature of Arrow's theorem. So Riker is correct: voting does not reveal our true preferences over all possible candidates (or ways of deciding an issue). But is this a Pyrrhic victory? If the result is used to proclaim the splendour of the market in relation to democracy, it most certainly is.

One of the normative claims of much of public choice theory – from the Virginia School more than the Rochester School, it must be said – is that the market is superior to democracy. The market is said to be efficient and it reveals the public's preferences for private goods, whereas democratic procedures are inefficient and, on Riker's argument at least, do not reveal the preferences of the electorate. But is this claim mere ideology? Why does voting get the bad press and the market such good press? Can we not apply Arrow's theorem to both?

On one interpretation Arrow's theorem is inapplicable to market processes since they are not designed to reveal collective preferences. We can easily see that few of us would choose the back-of-the-hand market outcomes we see all around us ‘depressions, speculative bubbles, involuntary unemployment, useless consumerism…. Dot-com entrepreneurs who waste other people's money are paid a thousand times what teachers or nurses are paid for taking care of human beings’.Footnote ⁴¹ But the market does not reveal our preferences for social states any more than do electoral processes. It only reveals our preferences given the constraints under which we buy and sell. If I buy a pint of beer rather than a bottle of champagne, does this reveal that I prefer beer to champagne? No. It shows I prefer beer to champagne, given my budget constraints. It reveals my preferences given the products on offer, and given my resources (which are measured relative to others’ resources). Similarly, voting reveals my preferences for the alternatives on offer given the constraints under which I vote. These include what candidates are standing for what policies, and the way I think others are going to vote. We operate strategically in markets, just as we do in elections.

Riker did not express what he meant by populism or liberalism very clearly. Given the title of his book, and the general thrust of the argument, that is a serious problem. But I take his underlying argument to be that we cannot expect democracy to reveal what the public ‘truly’ wants. As long as there is some correlation between getting rid of bad candidates and keeping good ones, then any democratic process is better than none. That is the liberal justification for democracy. In his final chapter Riker makes some claims that go well beyond what he has argued. Riker suggests the most liberal and best institutions are those contained in the US Constitution – but no one outside the United States takes that seriously. He suggests that the UK is the most populist of democracies. Why he says that is not clear, unless it is the fact that the British parliament is sovereign and so can override judicial decisions with new laws. Whilst Riker does draw some strange conclusions from his analysis, the analysis of democracy itself is well founded.

The real problem these days is not populist justifications of democracy, but populist justifications of the market. With complete information we can always try to counter any manipulation of decision rules with counter-manipulation. Without complete information, manipulation might bring advantage to some and loss to others; similarly in markets. There is a lot of manipulation in the market. If you put your house on the market at £320,000, but are prepared to accept £280,000 and I am prepared to pay £300,000 to buy it, offer you £260,000 and we finally agree a price of £290,000, then there has been manipulation. You tried to manipulate by setting a price higher than you were prepared to sell. And I offered a price lower than I was prepared to pay. What we revealed was what we were prepared to agree to under the bargaining constraints we were under. Now what I have described might be a ‘fair’ bargain. Given we agreed it, then, in some sense, it is an efficient one. But not all bargains are fair, and not all are efficient. If manipulation of this sort is OK in the liberal market, what's wrong with it under liberal democracy? In other words, our interpretation of democratic results needs to take into account the conditions under which choice is made. Similarly for the market. How we interpret social outcomes must take into account the constraints under which people operate. We need to take into account the information that people have, the incentives of business to act monopolistically, the regulation of markets created under the political constraints under which regulations are created, and the choices each of us make given the choices others are making.

I will give the last words to Gerry Mackie. He asks: why does current ideology present the results of democratic procedures pessimistically but the results of market procedures optimistically? Why does Riker think politics is about disequilibrium and economics about equilibrium? Arrow's first fundamental theorem of welfare economics is often used to promote the idea that the market is efficient and fair. But why not phrase it pessimistically?