Sources & method

Schedule 2, Table 1 of the Planning (Lease Variation Charges) Determination sets an additional dwelling amount per suburb, in eight bands keyed to the total number of dwellings approved after the variation. There is no API and no dataset for this — it is transcribed by hand from the determination on the ACT Legislation Register and should be verified there before you rely on any figure.

added    = max(0, proposed − existing)
band     = band_for(proposed)          # by TOTAL, not the increase
codified = added × rate(suburb, band)  # s.9(2)
payable  = codified × (1 − discount)   # if a concession applies

The codified schedule is calibrated to capture 75% of the value uplift. So:

implied_uplift        = codified ÷ 0.75   # NOT payable
your_share            = implied_uplift − payable
approved_land_value   = unimproved_value + implied_uplift
implied_approved_land = approved_land_value − payable    # the bid ceiling

The uplift is always derived from the codified charge, never the payable one. A remission changes what you pay; it does not change what the land is worth. Substituting one for the other roughly halves the implied land value and makes the screen produce nonsense.

The charge is then deducted, because you pay it. The two figures answer different questions and only one of them is a bid. The site is worth approved_land_value once the variation is granted — but you do not get there by buying the land, you get there by buying the land and paying the charge. So the most that can rationally be paid is the approved value less the charge payable, and that is what an asking price is tested against. Testing against the gross figure overstates the ceiling by the entire charge, always in the encouraging direction.

payable here, not codified: a concession you actually qualify for is money that stays with you, so it belongs in your ceiling.

ceiling = implied_approved_land         # approved value − charge payable

if    asking > ceiling:          WALK
elif  asking > ceiling × 0.85:   MARGINAL
else:                            WORTH MODELLING

Unimproved value is sourced from the ACT Open Data Portal, dataset rbu5-se5b, published per block for 2022–2026. Unit-titled parcels are published per unit; where no whole-of-block record exists the units are summed.

A $100,000 move in end value shifts residual land value by roughly $267,000, so this is the number worth arguing about — and the one the tool can least afford to state more confidently than it knows.

What is measured, and what is not. There is no published per-suburb sale price for completed dwellings in the ACT. The Land and Property Report publishes settlement bands for the whole Territory; the greenfield medians cover serviced land in nine new estates. So the estimate here separates two things that are usually conflated:

land per dwelling = median block UV in the division ÷ dwellings per block
raw               = land per dwelling + build cost
price per dwelling = raw × gross_up
price per m²      = price per dwelling ÷ living area

gross_up is solved once, so that the median division reproduces the Territory-wide figure. It absorbs margin, finance, selling and statutory costs — everything between "land plus build" and "what it sells for".

Why land value is not passed through directly. Median block value spans 4.4× across the Territory. Completed dwelling prices do not, because a dwelling is land plus a building and only the land part varies — the build costs much the same in Forrest as in Charnwood. Applying the land relativity unaltered would overstate the spread by the build share of the dwelling, which is most of it. The decomposition above damps 4.4× to about 1.9×.

Thin suburbs borrow from their district. A division's own median is blended toward its district by n ÷ (n + 60), so a suburb with 3,000 valuations stands on its own and one with 18 mostly does not. Each figure states which it is.

Every estimate carries a band — the block-to-block spread of land value inside that division — and a sentence saying what it rests on.

The level becomes measured one division at a time. Sales reported against a block feed back into this: once a cohort has enough that clear the strict gate — a completed new dwelling, an arms-length sale, corroborated rather than one person's word, and carrying a floor area — its median is blended toward the relativity by n ÷ (n + 8) and that division's level stops being an anchor. Six sales are better evidence than none and worse than 635,365 valuations, so the answer sits between them rather than replacing one with the other, and a cohort below the minimum is left alone entirely.

Both methods are live at once by design, so the caveat above appears exactly where the evidence does not. A licensed sample, when one arrives, supersedes both the same way.

revenue   = n × unit_price × (1 + market_drift)
gst       = max(0, (revenue − land) ÷ 11)        # margin scheme
duty      = land × duty_rate
build     = n × (living_sqm × liv_rate + garage_sqm × gar_rate)
statutory = lvc_payable + titling + fees
site      = demolition + siteworks + landscaping + consultants
conting   = (build + site) × contingency_pct
tdc       = land + duty + due_diligence + statutory + build + site + conting
debt      = tdc × lvr
interest  = debt × avg_drawn × rate × (months ÷ 12)
estab     = debt × estab_pct
selling   = revenue × agent_pct + marketing + sale_legals
itc       = (build + site + conting + selling + due_diligence) ÷ 11
total     = tdc + interest + estab + selling − itc
profit    = (revenue − gst) − total
margin    = profit ÷ total

Duty, finance and margin-scheme GST all scale with the land price, so the residual cannot be rearranged algebraically. It is solved by bisection over 80 iterations — margin falls monotonically as the land price rises, so there is exactly one root and the method is unconditionally stable.

The search floor is a land price of zero, and that matters. Margin is profit ÷ total cost and land is part of total cost, so at a sufficiently negative land price the denominator passes through zero and the margin flips sign. The function is only monotonic where total cost is positive. Zero is also the honest commercial floor: nobody is paid to take a development site, so where the target cannot be reached even with free land the tool says not viable at any price rather than reporting a negative number.

Cost inputs are GST-inclusive, and the GST comes back. Builders, consultants, demolition contractors and selling agents all quote inclusive, and the defaults were taken from quotes in that form. A developer selling under the margin scheme gives up the input tax credit on the land — that is the trade the scheme makes — but keeps it on everything bought to build and sell. So one eleventh of the build, site works, contingency, selling costs and due diligence is recoverable, and a model that ignored it would overstate cost by roughly a ninth of the build. On a three-dwelling scheme that is around $240,000, all of it flowing into an understated maximum bid.

Deliberately outside the credit: the Lease Variation Charge, titling and lodgement fees (government charges outside the GST net under Division 81), conveyance duty (a state tax, not a taxable supply), interest (financial supplies are input-taxed) and the land itself. Tick these costs are ex-GST on the calculator if you are working from ex-GST figures, and no credit is taken.

The credit is not netted off the finance base. In practice the GST is paid to suppliers and recovered on the next quarterly BAS, so it is genuinely carried — leaving it in the debt base is the conservative reading.

Establishment and exit fees are shown separately from interest. The source workbook this model was derived from multiplies that fee by zero, almost certainly in error. It is charged here at debt × estab_pct, which makes this model slightly more conservative than the spreadsheet. Set the rate to zero to reconcile the two.

Assumptions and their defaults