A good way to describe government is that it operates with rules that only one part of society, whether a majority or not, prefers to any other alternative, while another part would oppose them if it had the choice. These rules have predictability, efficacy and legitimacy, though not necessarily of some optimum measure. Where do they come from? Their sources are higher rules from which they can be deduced, and these higher rules will in turn be derived from yet higher rules in an infinite succession. In practical history, the succession has some end, like the founding father of a royal dynasty, a revolution, or the occupation of one country by another.
The succession of rules we are nowadays increasingly familiar with is the rule in which everybody chooses among one or more alternatives, and one alternative will come out of these choices as a valid rule. Binary choice by a simple majority is the simplest version of choice; it becomes a rule to which everybody becomes a subject even if he has chosen some other alternative that he preferred.
This rule, universally called democracy, has evolved in an almost straight line from early Church and Court hierarchies, nobility and property, until it took its final form with one man, one vote in late 19th century England and in countries following the English example, with one woman, one vote following it as an afterthought. Democracy has arrived not as the fruit of consideration and controversy, but almost as a moral axiom around which no supporting argument is necessary. Democracy, in effect, has arrived in its uncontested certainty rather like its broader version, equality, which has likewise occupied a rank of moral axiom from which practical moral rules have been deduced. No controversy and no reasoned argument existed about it. Nobody ventured to say that it would perhaps be reasonable to give higher voting rights to parents who are concerned about children than to bachelors who need only to worry about themselves. Such arguments would have probably been lost, but it is significant that they have not even been raised.
Democratic rule has some obvious arithmetical consequences. When people's wealth or incomes are unequal, the average is higher than the median, and redistributions from incomes above the average to below it will increase equality. With two voting blocks of equal size, the poorer side would win because it can seduce the marginal voters by offering them part of the income of the richer side. Redistribution can always defeat another distribution if it offers a higher transfer from incomes above the average.
Liberalism does not have any rule for a particular distribution of income to be decided by majority voting. In other words, it has no reward to offer for the size of the voting population whose distributing purpose is more attractive than the other side's.
As liberalism in its pure form has no distributive role in which its voting power decides between gainers and losers, rewards would be distributed by individual, and not aggregative, choices. Thus only distributive choices would take place in which the parties do not suffer losses, i.e. all sizes are voluntarily undertaken and not suffered by compulsion. Social choice has no place in this rule system. Imagine, for instance, two shepherds who have herds next to each other. Each has tried to steal sheep from the other, and each tries to guard his own sheep from the other. The stealing can only be successful if the other shepherd fails to guard his sheep. The interaction of the two is a net cost for each if they both steal or if they both guard their herd from the other. If they neither steal nor guard, they suffer no cost as long as the other behaves the same way. If neither decides to steal sheep nor decides to guard the sheep on a Sunday, that day will be more agreeable to each than the other day of the week. The "never on a Sunday" rule can experimentally be prolonged on Monday, and if it is again successful, on the following day as well, with each day adding to the cumulative gain of the two shepherds. The reciprocal practice is the more advantageous to each the longer it is practiced. The respect for the sheep of each shepherd becomes a conventional practice and a rule for respecting the property of the others. The rule, like any other that is effective, implies its own enforcement cost is born by deterrence of the damages that the participants of the convention would suffer if it was broken. Any reciprocal interaction in which the participants find an advantage to themselves (as long as an advantage occurs to the other party) is a Pareto-improving rule in principle. A behaviour rule which is enforced by the participants who derive an advantage for it will also endow its cost. The full set of such conventional rules, i.e. the protection of life and limb, property, and its exchange and the respect for freedoms in pursuit of peaceful purposes is a set of a liberal political order in which breaking the conventional order entails enforcing action by the participants. You might call this "order anarchy".
It is possible to imagine a liberal order in which assets and talents are given and goods and services are produced and distributed according to conventional rules, so that marginal rates of transformation and substitution tend to be equal. In this set of principles there is no place for any distributive role in which one part of society, relying on votes, would impose distributive shares on everybody, including gainers and losers.
Democracy can be imagined as having a rule-making rule in which certain rules are simply not admitted. Even if voting outcomes dictate that coloured people are not admitted at the front of buses or that women can only have jobs in which their conditions are worse that those that men obtain for the same jobs, such outcomes are not admitted because they are unconstitutional. Tax rule will have to be just and impartial. Nevertheless, taxes may be disproportional to income and be almost, though not quite, confiscatory. However, if this was not admitted, probably the most important function of voting for distributive outcomes would be prohibited and democracy would lose most of its greatest purpose, that is to say to bring about distributive outcomes that favour some at the expense of others. In such a case the rule-making rule will simply be used to amend itself like any other rule-making rule, and would permit the tax rule that has proven to be essential for a democratic purpose. In fact rule-making rules, i.e. constitutions, nearly always permit a great deal of freedom of taxation as well as lack of balance between government revenue and expenditure, because such options are permitted by the basic rules of social choice.
In pure liberal theory, or order anarchy, all rules are spontaneous conventions, which correspond to the individual preferences of the participants, who all profit from the convention (they are Pareto-improving). There is no collective will, and no expression of the collective will in a social choice rule. The distribution of the aggregate goods and services is a matter of property and contract among individuals.
When conventional rule is replaced by statutes rule, this order anarchy is overlaid, vetoed, modified, or completed by democracy. Democracy is operated by a social choice rule in which voting by one man-one vote is decisive (majority rule is usually required, though not necessary). The social choice rule decides the division of goods and services between two halves of society, one half gaining at the expense of the other.
As noted above, in a division between rich and poor halves of society the poor half would be the gainer, because it can get more value by gaining it from the rich than the rich could get by gaining it from the poor. In other words, in such a competition with other things equal, the poor dominate the rich in any redistribution. Simple arithmetic tells us that the social choice rule will be continuously busy, or at least busy from one election to the next, in its role as the redistributor of goods and services in favour of the poor at the expense of the rich.
Governments, for good reasons, will bring this about, not only by taxing the rich, but by their total expenditures in which the poor will make a great gain and the rich a moderate loss. The budget as a whole will keep expanding from one election to the next. This brings the deficit into a danger zone and forces the state toward austerity until the debt of the nation is reduced and fit again to continue the rich-to-poor redistribution. This may even continue to the point where the redistribution has achieved complete equality, with neither rich nor poor left. Redistribution at this point is still called for because otherwise the order anarchy in which complete equality is not a probable outcome would bring about a distribution and upset full equality. Democracy in this case would have to preserve the full equality that order anarchy would otherwise upset.
Despite the abstract nature that we had to describe the inevitable arithmetic confrontation between liberalism and democracy, the outcome is sufficiently discouraging to permit its translation into a more realistic and logically less strained picture. Individual preferences, liberties and probably also economic well-being appeared to be the victims and social choice is likely to be bought at the expense of prosperity. We cannot get much satisfaction from contemplating this likely picture. All we can get from the analysis of the forces at play is a better understanding of why liberal democracy is producing results which most of us most of the time do not welcome.