a/research· 14 October 2026 · 4 min read

Does a nation need one central authority to vote?

Art: Omxus Press

At about 4 am on 8 November 2023, a routine software change at Optus took its network down. About ten million Australians lost their phones. Trains stopped and hospitals lost lines. Over the next hours, 2,145 people who tried to call triple zero couldn't get through (Australian Communications and Media Authority, 2024).

Twenty-five days earlier, the same country had run the largest counting job it has. Nothing like that happened, because no single thing could break.

8,712 halls

On Saturday 14 October 2023, Australians answered one question. Nobody carried the ballots to Canberra. The votes were counted where they were cast: in 8,712 polling places and pre-poll centres, school halls and church halls and community centres, each sorting its own paper and reporting its totals upward. When the count was final, 15,895,231 votes had been recorded (Australian Electoral Commission, 2023). The national answer was a sum of local answers. If one hall's power fails, you get a late result from one hall.

We know how to build things that way. We've mostly stopped.

The network in the queue

Every voter in those queues carried a phone. Walking from the car park to the hall, it handed off between radio cells and renegotiated its signal many times a second. Even the phone companies have been learning the lesson of the hall. Their answer is called O-RAN, the Open Radio Access Network, and it splits the work by speed. A slow layer sets policy and keeps the inventory. A fast layer at the edge makes split-second decisions. Small plug-in apps do the specialist work. Every layer talks to the next through open, published interfaces, so no single company owns the stack (Polese et al., 2023). In 2018 two operator-led groups joined forces to standardise it, and within four years the alliance had more than 300 members.

Swap the towers for people

Nothing in that design cares what the things at the bottom are. Swap the towers for people and the coordination works the same way.

A person with a phone becomes the radio unit. Phones within Bluetooth range relay messages to each other, up to seven hops. The slow layer runs the calendar of national votes. The fast layer runs emergencies: someone calls for help, the alert spreads phone to phone, and the nearest verified person responds. When the towers go dark at 4 am, the street still works. The phone side of that comparison runs in commercial networks today. The civic side is a design under construction, not a working system. We're building it, so take our enthusiasm accordingly.

The usual objection is that civic life needs a central authority: no election without a central tabulator, no emergency response without a dispatch centre. But a phone network is harder than civic life on every count, and it is giving up its central controller anyway. The economist Elinor Ostrom found the same about human institutions: many centres of decision, nested and coordinating, govern complex systems better than a single state or a single market (Ostrom, 2010).

How heavy is a nation voting?

Take the hardest case, where every vote travels through a single computer as nine messages: the ballot, a receipt and seven relay hops. For everyone on the 2026 roll (Australian Electoral Commission, 2026), spread across a ten-hour day, that's about 4,600 events a second. We ran a small coordination program on two processor cores of ordinary hardware, and it handled more than 70,000 a second. That test measured how fast it receives and answers events, nothing more. But the sum makes the point: a nation voting is a light job. Four votes a year come to about 21 events a second, averaged out. The phone network does that before breakfast.

The two monkeys

In a famous experiment, two capuchin monkeys do the same task. One gets a grape, the other a slice of cucumber. The monkey paid in cucumber, seeing the grape, stops cooperating and throws the cucumber back (Brosnan & de Waal, 2003). The unfairness isn't complicated. The system handing out cucumbers is.

For most of the twentieth century that system could call itself a technical necessity. Someone had to run the database, count the votes, dispatch the ambulances. The defence weakened in 2018, when the operators of the world's phone networks committed to open, layered coordination. If they can split up a problem that hard, "civic life is too delicate" isn't a technical claim. It's a political one. Civic systems stay centralised because the parties running them gain from the centre. The centralised architecture isn't a requirement. It's a revenue model.

What you can do

Next election, work a booth or scrutineer. The Electoral Commission hires thousands of people, and you'll watch a nation get counted by people with pencils, which is one of the loveliest things we do. Keep a list of three neighbours' numbers on paper, with an agreed place to go if the phones die, because on 8 November 2023 a piece of paper beat the network.

And the next time someone says a service needs a single provider to run it, ask what happens to the people on it at 4 am when that provider breaks. Then ask how 8,712 halls manage.

Sources

Australian Communications and Media Authority. (2024). Optus penalised $12 million over triple zero failures in November 2023 outage [Media release].

Australian Electoral Commission. (2023). 2023 referendum: National results, votes by state and polling place results [Data set]. https://results.aec.gov.au/29581/Website/ReferendumNationalResults-29581.htm

Australian Electoral Commission. (2026). Enrolment statistics [Data set]. https://www.aec.gov.au/Enrolling_to_vote/Enrolment_stats/

Brosnan, S. F., & de Waal, F. B. M. (2003). Monkeys reject unequal pay. Nature, 425(6955), 297–299. https://doi.org/10.1038/nature01963

Ostrom, E. (2010). Beyond markets and states: Polycentric governance of complex economic systems. American Economic Review, 100(3), 641–672. https://doi.org/10.1257/aer.100.3.641

Polese, M., Bonati, L., D'Oro, S., Basagni, S., & Melodia, T. (2023). Understanding O-RAN: Architecture, interfaces, algorithms, security, and research challenges. IEEE Communications Surveys & Tutorials, 25(2), 1376–1411. https://doi.org/10.1109/COMST.2023.3239220

Keep reading