Single-Industry Dependence and Long-Term Capital Growth

Key Findings

• Mining-exposed suburbs trail the national median over horizons of twelve to twenty years. The median gap is two-and-a-half to four percentage points a year across those windows. Ninety-five per cent confidence intervals exclude zero at every window from twelve to twenty years. The twelve-year and fifteen-year windows show the largest gap, around four percentage points a year. This is a descriptive comparison against all other Australian suburbs, not a matched-controls causal estimate.
• Ten-year mining underperformance exists but is weaker and composition-sensitive. In the full panel the gap is about two-and-a-half percentage points a year. In a balanced panel restricted to the eighteen mining-exposed suburbs with complete twenty-three-year history, the ten-year gap shrinks to about one-and-a-quarter percentage points and loses statistical significance. The clear signal begins at twelve years.
• Short horizons depend on where you are in the commodity cycle. The five-year window ending 2026 shows no reliable gap because the current iron-ore rally has pulled performance back to parity. But rolling five-year windows ending 2008-2012 showed mining outperforming by five to twelve percentage points. Windows ending 2015-2020 showed underperformance of five to nine percentage points. The honest statement is that five-year mining performance tracks the commodity cycle, not a single “no-effect” conclusion.
• Twenty-three-year gap is small and not significant. Roughly half a percentage point a year, with a confidence interval that straddles zero. Whether this represents a genuine pre-boom-starting-price effect or a sample-attrition artefact cannot be determined with this data alone.
• Professional-services-heavy suburbs trail the national median on short-to-medium horizons. At the five-year window, professional-services-heavy suburbs (at least fifteen per cent of classified businesses in professional services, N=148) trailed by about four-and-a-third percentage points a year. At ten years, by about one-and-three-quarter percentage points. At fifteen and twenty years the gap is not reliably different from zero. These are mature destinations that lag cyclical national rallies but track the market over very long horizons.
• Dairy, wine, timber, fishing, fruit-irrigation and tourism suburbs do not show the mining pattern. Most specialist agricultural and lifestyle categories run at or ahead of the national median. This is reported as exploratory given the multiple-testing burden from ten categories tested across ten horizons We highlight the contrast between industrially concentrated single-sector exposure and diffused specialist exposure as a hypothesis worth further testing rather than a settled finding.

Background and Motivation

Conventional property-investing advice warns against towns dependent on a single industry. The warning is intuitive but rarely quantified. Australia has dozens of recognisable specialist-industry towns: mining towns, dairy regions, sugar-cane belts, smelter towns, tourism-dependent coastal strips, wine regions, fishing ports, timber towns, and wheat-and-sheep country. We tested whether the warning holds against the data, how large the effect is, and which specialist categories are actually exposed. The answer matters to cash-flow-seeking investors who tilt toward regional Australia for yield: some specialist tilts are free money, others are a slow-motion capital loss.

Methodology

We identify a suburb as mining-exposed when at least two per cent of its registered businesses carry mining-related names (mining, minerals, metals, exploration, pilbara, resources, ore, iron, quarry, drilling, and similar). Office-district contamination from capital-city central business districts (e.g. West Perth) is removed by excluding suburbs that sit within a Greater-capital area. We require at least fifty registered businesses per suburb to avoid over-reading on tiny samples.

The capital-growth comparison is median mining-exposed versus median rest-of-Australia, across time windows from two to twenty-three years, all ending in 2026. Each reported gap carries a ninety-five per cent confidence interval and a significance test. A finding is reported as significant only when both the confidence interval and the significance test agree on the direction.

A consistency check repeats every horizon analysis on only the mining-exposed suburbs with complete price history back to 2003. This eliminates the concern that the pattern is a sample-attrition artefact rather than a structural effect.

Separately, named specialist-town categories (dairy, sugar cane, timber, fishing, wheat-grain-sheep, wine, fruit irrigation, tourism, smelter, defence base) were tested against the national median. Because ten categories by ten horizons produces one hundred individual tests, individual specialist-category findings are reported as exploratory only and should not be treated as confirmed without independent replication.

Results

Mining-exposed suburbs versus the national median

Every row is the median gap in percentage points per year, with a ninety-five per cent confidence interval and a significance decision. Horizons where the gap is not reliably different from zero are explicit.

Why so few mining towns? Australia has about forty named mining-dependent towns. Many are too small to produce a reliable annual median house price. The sample includes every mining-exposed suburb with enough sales to compute a twenty-three-year price history. These are Karratha, Newman, Kalgoorlie, Boulder, Port Hedland, South Hedland, Roebourne, Dampier, Tom Price, Moranbah, Dysart, Blackwater, Middlemount, Clermont, Emerald, Moura, Cloncurry, Mount Isa, Lightning Ridge, Coober Pedy, Andamooka, Roxby Downs, Kambalda West, Kambalda East, Coolgardie, Leonora, Norseman, Tennant Creek, Glenden, and similar towns. The sample is small because mining settlements are small, not because the classification misses them.

We also repeated the analysis on only the eighteen mining-exposed suburbs with complete price history back to 2003. The twelve to twenty year gap survives this check at similar magnitudes. The ten-year effect weakens substantially on the same eighteen-suburb set, which is why we do not report a consistent ten-year finding.

Specialist-category results (exploratory)

Separately from the mining-exposed classification, we tested ten named specialist-town categories against the national median. These results should be treated as exploratory because the ten-category by ten-horizon testing matrix carries a substantial multiple-testing burden that we have not formally corrected for. No individual specialist-category finding from this table should be cited without independent replication.

Mining-exposed is the only category where we also repeated the analysis on a data-rich subset. Smelter and heavy industry at sample size six does not pass our reliability bar. We do not include it in the reliable set. Sugar cane and wheat-grain-sheep show tentative long-horizon underperformance but would not survive a multiple-testing correction across all the categories tested.

Illustrative examples

Two suburbs illustrate the contrast between concentrated industrially concentrated exposure and diffused specialist exposure.

Karratha (Regional Western Australia). A classic single-industry mining suburb in the Pilbara. Its local business composition is dominated by trades and support services for the iron-ore workforce. Median house price peaked in 2017 and fell sharply through 2018-2021. Its annualised capital growth over the fifteen years to 2026 trailed the national median materially. The cash-flow yield during the same window was well above the national average, partly offsetting the capital shortfall, but a buyer who purchased for yield and implicit capital appreciation received the yield and not the appreciation.

Leongatha (Gippsland, Victoria). A dairy-industry suburb spanning many small operators with no single large employer. Its annualised capital growth over the decade to 2026 was approximately in line with the national median. The specialist-industry label is identical in category (both Karratha and Leongatha are regional and single-sector dependent), but the concentration structure is opposite. One large employer in Karratha, many small ones in Leongatha.

Professional-services-heavy suburbs

We also tested a broader cut: every Australian suburb where professional services (lawyers, accountants, financial planners, consultants, architects, management consultancies) make up at least fifteen per cent of the classified local business mix. This captures the Toorak, Mosman, Paddington and Subiaco pattern without requiring professional services to be the single biggest category.

Professional-services-heavy suburbs underperform reliably over five and ten year horizons. The effect fades at fifteen and twenty years. Interpretation: these are mature wealth-preservation destinations. They do not participate in cyclical national rallies, which is why they lag on shorter horizons. Over the very long run, they track the market. A buyer hoping for capital growth over five to ten years in one of these suburbs is likely to be disappointed.

The time dimension, looked at honestly

The single-endpoint analysis above ends every horizon in 2026. That is a design choice that privileges the current iron-ore rally. A cleaner test is rolling-origin: compute the five-year, ten-year and fifteen-year median gap ending in every year since 2008, not just the year 2026. That reveals how the mining-exposed gap moves with the commodity cycle.

The rolling analysis tells a boom-bust story, not a long-run underperformance story.

• 2008-2012. Five-year rolling windows ending in these years showed mining-exposed suburbs outperforming the rest of Australia by five to twelve percentage points a year. This was the commodity super-cycle.
• 2013-2014. The five-year gap collapsed to near zero as the peak passed.
• 2015-2021. The five-year gap turned sharply negative, reaching minus-nine percentage points a year at the 2018 end point. This is the post-boom capital decline.
• 2022-2026. Five-year windows ending in these years show no reliable gap. The current iron-ore rally has pulled rolling performance back to parity.

On the ten-year rolling window, the pattern is similar but dampened. Ten-year windows ending 2016 through 2026 all show significant underperformance, ranging from minus-two-and-a-half percentage points at the most recent endpoint to minus-eight percentage points at 2017. On the fifteen-year rolling window, every end year from 2018 to 2026 shows significant underperformance of two to five-and-a-half percentage points a year.

The long-horizon “mining trails the market” finding holds because every fifteen-year window that contains the 2013-2020 bust captures real underperformance. It does not hold because mining suburbs are structurally slow growers. During the boom they were the fastest growers in the country. The honest summary is that mining property follows the commodity cycle with a multi-year lag.

Matched-controls check

A legitimate objection is that the comparison group “all other Australian suburbs” mixes metro, coastal and growth-corridor suburbs against remote mining towns. We rebuilt the comparison by matching each mining-exposed suburb to up to five regional non-mining suburbs in the same state and the same starting-price quintile. Matched comparison at selected end years and horizons:

The matched analysis is the more honest benchmark and confirms the direction of the finding but at about half the magnitude of the unmatched comparison. Matched ten-year gaps are minus-one to minus-three-and-a-half percentage points a year, versus minus-two-and-a-half to minus-seven in the unmatched version. The fifteen-year matched gap of minus-three-and-a-half at 2026 remains significant. This tells us roughly half the raw underperformance was remoteness and starting-price composition. The remainder is specific to mining exposure.

Defense Against Criticism

The short windows show no gap ending 2026, but that is cycle-dependent.

We addressed this by rolling the five-year window back to every end year from 2008. The gap swings from strong outperformance during the 2003-2012 boom to severe underperformance during the 2013-2020 bust, then back to parity during the current rally. The five-year window ending 2026 shows no gap because the iron ore price is high. The five-year window ending 2018 showed minus-nine percentage points a year. A buyer with a five-year hold cannot know in advance which side of the cycle they are on.

The mining-exposed sample is small and perhaps non-representative.

The data-driven classification identifies between fifteen and twenty-nine suburbs at each horizon. The finding survives the sample-size concern in two ways. First, the confidence intervals exclude zero at every window from twelve to twenty years. Second, repeating the analysis on just the fifteen suburbs with full twenty-three-year history reproduces the same direction and magnitude at twelve to twenty years. The result does not depend on which mining-exposed suburbs are included at the margin.

All-other-suburbs is not a matched control group.

We have now run the matched comparison. Each mining-exposed suburb was matched to up to five regional non-mining suburbs in the same state and the same starting-price quintile. The matched ten-year gap is minus one to minus three-and-a-half percentage points per year depending on end year (unmatched was minus two-and-a-half to minus seven). The fifteen-year matched gap at the 2026 endpoint is minus three-and-a-half percentage points a year and remains significant. About half the raw underperformance in the unmatched analysis was remoteness and starting-price composition. The remaining half is specific to mining exposure. See the matched-controls table in the Results section.

Specialist-category findings are exploratory.

We ran ten specialist-town categories against the national median across ten horizons, producing one hundred individual comparisons. With that many tests, roughly five false positives are expected by chance alone. We therefore do not report any individual specialist-category finding as confirmed. The mining-exposure finding is reported with confidence because its classification was defined by a single rule before horizon analysis and because it holds up on the smaller, more data-rich subset of mining towns.

What this study does NOT prove

The paper has been through adversarial review and several methodological limits are worth stating plainly before the conclusion.

• Not causal. This is a descriptive comparison of mining-token-heavy suburbs versus all other Australian suburbs. No matched controls, no statistical modeling with covariates, no instrumental variable. The observed gap could reflect remoteness, state, starting-price composition, or liquidity rather than mining exposure itself.
• Not forward-looking. The mining-exposure classification uses the 2023 business register to analyse growth windows starting in 2003 onwards. That is after-the-fact classification, not a leading-indicator test. A forecast test would require using an earlier snapshot to predict a later window. We have the 2017 snapshot and have not yet run that test.
• Not independent observations. The twenty-six to thirty-one mining-exposed suburbs are concentrated in a handful of commodity regions (Pilbara, Goldfields, Bowen Basin, Mount Isa). Standard statistical tests treat them as independent. They are not. The confidence intervals and p-values overstate precision.
• Not ten independent horizon tests. Every horizon ends in 2026 and shares the same endpoint. Nested horizon tests are not independent.
• Not a measure of single-employer exposure. The treatment is the share of business names carrying mining-related tokens. It is not mining employment, mine proximity, or workforce concentration. We call it mining-token-exposed, not single-employer-exposed.
• Noisy small-town medians. Annual median prices for small mining towns can move on three sales. We have not imposed a transaction-count floor.

Limitations

This paper addresses capital growth only. Rental yield in mining-exposed suburbs is materially above the national average, and a total-return analysis would narrow but not eliminate the capital gap at long horizons. The classification is a business-name-based proxy. Registered business names are noisy. A more rigorous classification would combine name-token exposure with proximity-to-mine data and employment-by-industry data. The national-median benchmark has not been adjusted for state, remoteness, or starting-price composition. Specialist-category findings beyond the mining-exposed group should be treated as exploratory and not cited without independent replication.

Conclusion

Australian mining-exposed regional suburbs, identified by the share of their registered businesses carrying mining-related names, show consistent long-horizon capital-growth underperformance against the national median over twelve to twenty year windows ending in 2026. The median gap is two-and-a-half to four percentage points per year, significant at ninety-five per cent confidence, and consistent when the analysis is repeated on only the mining towns with the longest price histories. Short horizons show no reliable gap. The business composition is visible in real time and is a practical risk-flag for investors considering long holds in regional suburbs.

Beyond the mining-exposed group, we report specialist-category results as exploratory. The substantive hypothesis worth further testing is that the underperformance pattern is specific to industrially concentrated single-sector exposure rather than to specialist regional economies generally.