Parchment and the Crowd: The Fourth Condition of a Just Democracy

A constitution's promises tell you least about it. What tells you most is what a majority must do to take one back, and whether anyone can take it without asking.

Losing an election has to be survivable. A right the winners can remove is a loan.

The Just Democracy framework sets four conditions that a vote cannot supply and cannot substitute for. The fourth is a state with the institutional capacity to protect individual rights against the will of organized majorities. Capacity here is about routes, not resolve. A right is a question taken off the ballot. This condition asks two things about any such question: what a majority must do to put it back, and whether anyone can settle it without a ballot at all. The first is answered in the statute book, where the price is an ordinary bill, a supermajority, a referendum, or no lawful route at all. The second is answered on the street. A protection that survives the first and fails the second was never protection, only paperwork.

The condition is not a list of rights or judicial review, though a court is often the obstacle. It does not say what must stand in the way, only that something must, and it leaves almost every question to the ordinary process. A high price of removal is not a virtue in itself. It protects whatever it covers, and a constitution can make an exclusion as hard to undo as a right. The condition asks about individual rights, not about how hard a constitution is to amend.

Two terms need pinning down. An organized majority is any coalition able to impose its will on a question if nothing stands in its way, whether it acts through a legislature, a referendum, or a crowd. It need not be a numerical majority, only the side that would win. An individual right is a claim a person holds against the state, one the state is bound to defend against every such coalition. That is why a crowd can take it while every law protecting it still stands. That second sense is a premise rather than a finding, and the second question rests on it.

James Madison doubted that written guarantees would help. To Jefferson in October 1788 he wrote: "Repeated violations of these parchment barriers have been committed by overbearing majorities in every State." He concluded that the danger of oppression lies wherever real power lies, which in a popular government means the majority. He drafted the Bill of Rights anyway. His doubt survives as this condition's first question. A guarantee is worth what it costs to remove.

Islamic jurisprudence took up the first question in its own terms. Classical jurists distinguished the rights of God, which the ruler enforced for the community, from the rights of persons, which belonged to the individual who held them. The latter could be waived only by their holder.

Jews, Christians, and other peoples of scripture lived under a pact, the dhimma, which secured their lives, property, and worship. In return, free adult men paid a poll tax, the jizya. Once it was paid, al-Mawardi held, the ruler was bound to protect those under the pact. When the Mongols overran Damascus in 1300, the jurist Ibn Taymiyya, by his own account, insisted that they free the Jewish and Christian captives along with the Muslims, as people under Muslim protection. The pact also carried a lower legal status, which the tradition's defenders read today as the custom of its age and its critics as subjugation. The price of keeping it fell on the minority it protected.

The second question has a court record. In 1988 the European Court of Human Rights heard Austrian doctors whose marches against abortion had been disrupted by counterdemonstrators. It held that a state must take reasonable steps so that people can demonstrate without fear of violence from their opponents, and it found that the Austrian police had not failed in that duty. In 2012 a large crowd from two Orthodox groups attacked about thirty LGBT activists in Tbilisi while the police failed to protect them. Three years later the Court applied the same rule and found Georgia in breach. One rule stood behind both causes. The evidence the second question asks for is what the law does when a crowd threatens a lawful exercise of a right: restrain the crowd, or restrict the right.

Most constitutions have a lock, but not every lock has a key. In Canada the key hangs on the majority's side of the door. Any legislature may declare that a law operates notwithstanding the Charter's guarantees of fundamental freedoms, legal rights, and equality. The Supreme Court ruled in 1988 that the notwithstanding clause asks only for form, and that a legislature need not justify the choice or show that any right was infringed. In practice legislatures have invoked it before any court has ruled. Setting any of those guarantees aside, for any law a legislature picks, therefore costs one line in an ordinary bill. The declaration lapses after five years, and the text sets no limit on renewal. An election normally comes before a renewal, so the question returns to the ballot bundled with everything else a general election decides.

Quebec invoked the clause for its 2019 laicity law, which bars many public servants from wearing religious symbols at work, and renewed it in 2024. In March 2026 the Supreme Court heard the challenge, including whether a legislature may use the clause before any court has ruled. Judgment is pending, and Quebec did not wait for it. In April 2026 the province extended many of those restrictions to postsecondary institutions and shielded the new law with the clause in advance.

The clause stops short, and where it stops Canada meets the condition. It cannot reach the right to vote, the right to move and live anywhere in the country, or minority language schooling. Removing those takes a constitutional amendment. This series has credited Canadian amendability before, on the ground that Parliament can repeal what Parliament passed. Both claims hold. Amendability is a virtue where what can be amended is a bad statute, and a defect where it is a rights protection.

The second question has gone to court too. After protests, some of them violent, outside synagogues and a Hindu temple in 2024, cities from Vaughan to Ottawa barred demonstrations near places of worship. The Canadian Civil Liberties Association has challenged Vaughan's ban in court as a limit on expression and assembly.

Switzerland puts rights on the ballot by design, and the Swiss vote several times a year, so a referendum is the ordinary process rather than a higher price. In 2009, 57.5% of voters and a majority of the cantons wrote a ban on building minarets into the federal constitution, over the opposition of the government and Parliament. No Swiss court can strike the ban, because it is in the constitution. Removing it now takes another vote of the people and the cantons, the same price as removing any right in the constitution. The only substantive limit on an initiative is the peremptory core of international law, a rule the Swiss did not write. The idea that a people's will has limits is older than that rule. Benjamin Constant argued in 1815 that popular sovereignty is bounded by justice and by the rights of individuals: "The will of a whole people cannot make just what is unjust."

Slovenia answered the same question the other way and wrote the answer down. Since 2013 its constitution has forbidden a referendum on any law that corrects what the Constitutional Court has found unconstitutional. The bar did not stop voters from rejecting a same sex marriage law by nearly two thirds in 2015, because no court had yet ruled on marriage. In 2022 the court held that excluding same sex couples from marriage and adoption was discrimination. Parliament amended the family code, opponents demanded a referendum, Parliament declared it inadmissible, and the court agreed. The concession is in the timing: Slovenia's protection closes the ballot only after judges have spoken.

None of this belongs to one side. The Swiss and Slovenian majorities voted from the right. Mexico's voted from the left. In 2024 its governing coalition reached a two thirds majority in the Senate and made the judiciary elected. Within weeks it amended the constitution to put amendments beyond challenge. When the Supreme Court took up a challenge to the reform, striking any part of it required eight of eleven votes, and the challenge fell short. A supermajority protects only against a majority that does not hold it.

Pakistan answers the first question with a supermajority, and the second with a crowd. In 1974 Parliament amended the constitution to declare Ahmadis not Muslim for the purposes of law. A 1984 ordinance made it a crime, punishable by up to three years in prison, for an Ahmadi to call himself a Muslim or his faith Islam. What Pakistan took off the ballot in 1974 was not a right but its denial, and putting the question back takes another supermajority. In 1993 Pakistan's Supreme Court upheld the ordinance, reasoning in part that Ahmadis' public use of Islamic terms and rites could provoke violence and so was not protected. In effect, the court let the crowd's anger settle the right before the crowd arrived.

The crowd came anyway. The Centre for Social Justice, in an April 2025 report, counts at least 104 people killed outside any legal process after blasphemy accusations between 1994 and 2024, 64% of them Muslims. The Muslims among the dead illustrate what an organized majority is. It is not the religious majority but the side that would win, and its reach includes the majority's own members. In 2011 a provincial governor was shot by his own guard over the blasphemy law, and the federal minister for minorities was killed later that year. A formal protection can hold while the person it protects loses the right.

Jeremy Waldron has put the case against judicial review of legislation at its strongest. His case assumes a society with working democratic institutions and a public committed to rights, even while it disagrees about what they are. In such a society, he argues, there is no reason to think judges protect rights better than legislatures, and strong judicial review is democratically illegitimate whatever its results. Canada is that society, and its notwithstanding clause gives the legislature the last word. Waldron, writing when legislatures rarely used it, still counted Canada's arrangement as strong review. The clause's defenders say it carries a price, paid at the next election. But a price collected only when the public objects is one the majority sets for itself, and in Quebec the people who bore it were not in that majority.

Slovenia's protection runs through a court, and there the objection lands hardest. Judges overruled a definition voters had defended at the ballot, and some defenders of popular sovereignty will think that cost too high. But Waldron's case, as he states it, holds only where the public takes rights seriously. Where it does not, the objection lapses. In Pakistan the laws and the killings have answered that question, and Waldron's case does not reach a state where they have.

Which questions belong off the ballot is the harder question, one this condition does not answer. It asks, as the third condition did, what removing a protection costs, not which protections to have. A state that takes the wrong claim for a right will find the mistake expensive to correct, and that expense is the condition's price.

An election can settle more than it should when law bends for the powerful, when arms answer to no civilian, when citizens cannot find out what is done in their name, and when the winners can remove a right. Every protection in this essay began on parchment. Whether it stayed a right depended on what a majority had to do to take it back, and on who stood between the crowd and the right when the crowd came.