The Machines Find the Poorest Streets

Every three months, the New South Wales gaming regulator publishes a spreadsheet with little fanfare. It records how much money was fed into poker machines in every corner of the state — club by club, hotel by hotel, council by council. It is dry, and it is enormous: in the year to mid-2025, New South Wales lost a record $9.1 billion to the machines, more than any state in the country and a sum that would rank the state among the heaviest per-capita gambling losses on earth.

Read the spreadsheet closely, and set it beside the Bureau of Statistics’ map of socio-economic disadvantage, and it begins to describe something the numbers were never designed to say out loud. The machines are not spread evenly. They gather — physically, in the count of devices per resident — in the suburbs with the least to spare. And where most of the state’s people live, the money follows them.

The pattern is real. It is also more conditional, and more interesting, than the slogan version. This is what six years of the state’s own records show.

Where the machines are#

Start with the machines themselves, because that is where the evidence is cleanest. The Audit Office of New South Wales — not an advocacy group, but the state’s own auditor — went looking in 2025 for whether gaming-machine regulation protects the people most at risk. It found the opposite of even distribution.

In Fairfield there is one poker machine for every 55 residents. In Ku-ring-gai, there is one for every 1,443.

Audit Office of New South Wales, June 2025

Fairfield, in Sydney’s west, is the most disadvantaged local government area in Greater Sydney on the Bureau’s index. Ku-ring-gai, on the leafy North Shore, is among the most advantaged. The auditor found the three council areas with the densest concentration of machines — Fairfield, Canterbury-Bankstown and Cumberland — all rank among the city’s most disadvantaged, and concluded that the regulator’s strategy “does not have a sufficient focus” on high-risk areas.

The regulator’s per-council data bears the auditor out. Set the machine count in each area against its population, across a clean sample of 49 council areas — the ones that report both club and hotel figures for a single council, together covering 82% of the state’s losses — and the density of machines and the depth of disadvantage move together closely (a rank correlation of −0.51, where a perfectly inverse relationship would be −1). The poorer the area, the more machines per adult. Decile by decile, machine density falls almost in a straight line, from about 23 machines per thousand adults in the most disadvantaged areas to roughly 11 in the most advantaged.

This is a supply story as much as a demand one. The machines are placed; someone decided where. And they are placed, disproportionately, where disadvantage is deepest.

The losses follow — but mind the weighting#

Do the dollars lost per adult track disadvantage as tightly as the machines do? Yes, but less cleanly — and only in the parts of the state where most people live.

Rank the 49 councils by disadvantage and by loss per adult, and the plain correlation is −0.21 — the right direction, but weak, and short of the −0.4 threshold set in advance, before the numbers were run. Weight each council by its adult population, and the relationship strengthens sharply, to −0.46. That gap matters. The councils that fight the trend are mostly small regional ones, which scatter widely; the large, populous, disadvantaged councils of Sydney’s west — Fairfield, Canterbury-Bankstown, Cumberland, Liverpool, Campbelltown — are heavy losers almost without exception. Where the people are, the pattern holds.

Split the sample by geography and the point becomes unmistakable. Within Greater Sydney, the link between disadvantage and losses is strong (−0.70). Across regional New South Wales, it vanishes entirely (+0.03). In the country, per-adult losses are high more or less regardless of disadvantage — a product of fewer venues serving larger catchments, of tourism, and of towns on the Victorian border where the machines draw players from across the river. This is, first and foremost, a metropolitan phenomenon.

Fairfield sits at the extreme. In 2024–25 its adults lost an average of $4,255 each to the machines — the highest in the sample, about 3.6 times the sample median of $1,195, and many times what adults lose in the advantaged councils of the northern beaches or the southern highlands.

Not a tidy staircase#

It would be neat to report that losses climb smoothly from the richest decile to the poorest. They do not.

The most disadvantaged decile in the sample is a single council — Fairfield — which makes it an outlier, not a category. The middle deciles are a jumble. And the most advantaged decile actually loses relatively little per adult, because it contains places like the northern beaches and the inner west where, whatever else is true, people are not feeding the rent into a machine. Taken as three broad bands, the most disadvantaged third of councils lose about 1.75 times as much per adult as the most advantaged third — a real gap, but not the clean doubling the simplest version of the story would predict.

Over six years, that concentration has barely shifted. The share of losses falling on the most disadvantaged areas has hovered around 35–36% throughout, even as the total rebounded hard from the pandemic’s venue closures.

The state’s cut#

There is a second party to every dollar lost: the government. New South Wales taxes gaming-machine profits, and as losses have climbed, so has the state’s take — from $1.38 billion in 2019–20 to $2.53 billion in 2024–25, an effective rate that has drifted up toward 28 cents in every dollar lost.

That is the uncomfortable arithmetic of reform. A government that moved decisively to cut losses would be cutting its own revenue at the same time — revenue that, as the machines have grown more lucrative, the budget has come to lean on.

Following the money, carefully#

The lazy version of this story says the gambling lobby buys its protection with political donations. The truth is more tangled, and getting it wrong would be a gift to the industry.

New South Wales has, unusually, banned political donations from the gambling industry — but the ban arrived in two stages, and the gap between them matters. Since late 2010, for-profit gambling businesses — hotels with poker machines, casinos, machine manufacturers — have been “prohibited donors” at the state level. But the definition turned on making a profit, and the not-for-profit registered clubs that run most of the state’s machines fell outside it. Clubs could lawfully donate to state campaigns until June 2023, when the incoming Minns government closed the loophole. So the tidy claim that “gambling money is banned in NSW” was, until very recently, only half true.

Influence, in any case, was never mainly a matter of state cheques. It ran — and runs — through other channels: federal donations, which no such ban touches; third-party advertising campaigns; and the memoranda of understanding that ClubsNSW signed with parties before elections, trading electoral support for policy commitments.

At the federal level the money is real but modest by the standards of corporate Australia: gambling-industry entities have given the major parties about $24 million across all the years the Electoral Commission has disclosed, with the Australian Hotels Association the single largest source. In the most recent year the flow tilted toward Labor. Set against a nine-billion-dollar loss pool, it is less a purchase than a retainer — enough to keep the door open, on both sides.

Reform, stalled#

The case for reform is not new, and neither is its defeat. In 2010 the Productivity Commission recommended $1 maximum bets and binding pre-commitment limits; a deal to enact them, struck between the independent Andrew Wilkie and the Gillard government, collapsed in 2012 after an industry advertising campaign. In 2022 the NSW Crime Commission’s Project Islington inquiry found that a significant share of the tens of billions cycled through the state’s machines was likely the proceeds of crime, and recommended mandatory cashless play. A cashless trial followed in 2024. In November 2024 an independent expert panel delivered a 530-page Roadmap for Gaming Reform with 30 recommendations.

More than 500 days later, the government has not formally responded. Losses, in the meantime, have set a record. The machines are still being placed.

What could explain this instead#

A correlation between disadvantage and gambling is not proof that the machines cause the hardship. There are at least four alternatives that fit the same data. Supply follows cheap land, not poverty. Clubs and hotels were sited decades ago where land was affordable and foot traffic reliable — which often means working-class suburbs. On this reading the machines are dense in disadvantaged areas because that is where the venues historically are, not because operators target the vulnerable. The two are hard to separate with council-level data alone. Reverse causation. The disadvantage index is built partly from income and unemployment — the very things gambling losses can worsen. Some of the correlation may run from harm to “disadvantage,” not the other way. Ecological inference. These are council averages, not people. An area can be disadvantaged on average while its heaviest losers are not its poorest residents. Council-level patterns cannot, on their own, establish who is losing the money. The weighting itself. The losses correlation only clears the threshold when the data are weighted by population, and it is carried by a handful of large western-Sydney councils. That is a real feature of where harm concentrates — but it means the result rests on relatively few high-leverage places, and a different, defensible choice of method would report a weaker link.

None of these dissolve the finding. The machine-density gradient is robust, the auditor’s ratios are stark, and where most of the state’s people live the losses track disadvantage. But the evidence supports the descriptive claim — the machines cluster in the poorest streets — more firmly than the causal one.