Agree to Disagree: A Short History of Coalition Partners in Public

In a single week, Winston Peters accused the Government he belongs to of covertly changing migration settings, and ACT publicly questioned the timing of National’s flagship parental leave announcement. In most parliamentary democracies this would read as a government coming apart. In New Zealand it reads as a Tuesday, and the difference is worth understanding, because it is built into the system rather than a symptom of its failure.

The old rule was collective responsibility: once Cabinet decides, every minister defends the decision in public, whatever they argued in the room. It is a Westminster inheritance, and under first-past-the-post it worked well enough, because governments were single parties and dissent could be handled at caucus. Then MMP arrived in 1996, coalitions became the normal condition of government, and the rule met a problem. Small parties that spend three years agreeing with a larger partner tend to be absorbed or annihilated at the next election. New Zealand First and ACT have both lived this. Total unity, for a minor coalition partner, is a slow form of suicide.

So the system adapted, and the adaptation has a name. The Cabinet Manual now provides for “agree to disagree” processes, under which coalition parties can maintain distinct positions on identified issues while still supporting the government on confidence and supply. It is collective responsibility with designated exits. A party can vote for the Budget in the afternoon and distance itself from a policy in the evening, provided the distancing happens in the agreed lane.

What this week showed is how much traffic the lanes now carry, especially with an election approaching. A coalition partner attacking the Government’s transparency is not trying to bring the Government down. He is differentiating a product. ACT questioning National’s announcement is doing the same. Each party needs its own voters to believe it fought for something specific, and the fighting must be visible to count. The performance is real in the sense that the disagreements are genuine. It is theatrical in the sense that everyone involved intends to be in coalition again if the numbers allow.

The practical guide for the observer is therefore about thresholds. Public sniping between coalition partners is the system working as designed and signifies little. The signals that actually matter are votes: a partner threatening confidence and supply, or refusing to vote for legislation the coalition agreement covers. Short of that, what you are watching is not a government in crisis. It is three parties sharing one government while campaigning for three different futures, out loud, on schedule.