The panel spread — why the same file gets two different answers.

Borrowing capacity is the number a borrower wants to know before they speak to anyone — agent, broker, parents, themselves. Most calculators on most websites return a single figure. Whoever runs the calculation owns the answer. The model assumed by that interface is: a borrower has a maximum, the calculator estimates it, and that estimate is roughly the same regardless of where you ask.

This model is wrong. Borrowing capacity doesn't exist as a single number for a single borrower. It exists as a distribution across the panel of lenders that might write the loan. Same income, same deposit, same liabilities, same credit file — the panel will return a borrowing-capacity range, not a point. And the range is wider than most borrowers realise: routinely 20-40% between the most-generous lender and the most-conservative lender on the same shaped file.

The reason this matters: the panel spread isn't broker performance variance. Two competent brokers running the same file across the same panel of 98 active home-loan lenders will get the same spread, because the spread is produced by the lenders' policies, not by who's asking. What varies is which lender each broker chooses to lead the application with — and whether the borrower understands the spread exists at all.

The four structural levers behind the spread

The panel spread comes from differences in how lenders apply four specific policy levers. Each one is set at the lender's discretion, ratified by their credit committee, and updated periodically as the lender re-prices risk. None of the four is set by APRA or the RBA; they're competitive policy variables.

1. The assessment rate

Lenders don't assess a borrower's serviceability at the rate offered. They assess at a higher "assessment rate" — the rate the borrower would be paying if interest rates rose. APRA requires the assessment rate to be at least the offered rate plus 3% buffer. Practically every lender sits between 8.0% and 9.0% on assessment rate in mid-2026, but the spread within that band is meaningful.

A $1M loan assessed at 8.30% requires roughly $7,800/month in serviceable income. The same loan assessed at 8.95% requires $8,250/month — about 6% more. The lender at 8.30% will service borrowers who fall short at 8.95%. The spread isn't dramatic on its own, but it stacks with the others.

2. The DTI cap

Debt-to-income ratio (DTI) is total debt divided by gross income. Most APRA-regulated banks cap DTI at 6.0x. Some go to 7.0x. Some non-bank lenders go to 8.0x or 9.0x. On a $130,000 income, a 6x DTI caps total debt at $780,000. A 7x DTI caps at $910,000. A 9x cap allows $1,170,000.

The DTI cap variance is more material than the assessment-rate variance because it operates as a hard ceiling — once you hit it, the lender stops, regardless of how strong the rest of the file looks. A borrower with strong income and weak ancillary debt might find that an APRA-bank 6x cap simply turns the file off, while a non-bank 8x cap waves it through to a much larger loan amount.

3. Income shading methodology

Lenders apply varying "shading" rules to non-base income components. Overtime, bonus, penalty rates, commission, rental income, allowances — each component gets reduced before being counted toward serviceability. The shading rules are where the panel diverges most sharply.

• Rental income is shaded 75-85% across the panel. NAB's 85% sits at the high end; a few non-banks use 70-75%.
• Bonus and commission are shaded 60-100% depending on consistency history and lender.
• Self-employed income is averaged differently lender-by-lender — lower-of-two-years, latest-capped, most-recent-shaded, or straight-average. The same returns produce assessable-income numbers $70-150k apart across these methodologies.
• Penalty rates and shift loadings for nurses, police and shift-workers are shaded 80-100%. Police-friendly lenders count them at 100%; mainstream lenders shade at 80%.

Income shading is where the broker file shape compounds with lender selection. A self-employed file with two strong years gets lender choice 1; a self-employed file with one strong year and one weak year gets a completely different lender choice — even though both files have the same trailing two-year average.

4. The HEM benchmark and how living expenses get calculated

Lenders use the Household Expenditure Measure (HEM) as the minimum living-expense assumption. They then compare HEM against the borrower's declared expenses and use the higher of the two. The variance comes in how each lender adjusts HEM for postcode, household size, dependents and income tier. Some lenders use a stricter HEM uplift schedule than others. On a high-income file with no real expense discipline, the HEM number can move serviceability by $50,000-$100,000 between lenders.

Why the panel spread doesn't close over time

An obvious question: if some lenders are generous and others are conservative on identical files, why doesn't the market converge? Why doesn't the cheap lender absorb all the volume and force the others to match?

The answer is risk pricing. The generous lender is generous because they're pricing for the marginal risk that comes with it — wider DTI caps mean their book takes more leverage; higher rental shading means their book carries more vacancy exposure; relaxed self-employed methodology means their book takes more income-volatility risk. The conservative lender is conservative because they want a cleaner book and they're willing to take less volume to get it.

Both positions are commercially rational. They just produce different answers for the same borrower. The market doesn't converge because the lenders aren't trying to converge — they're competing on different risk axes, not on a single price.

The panel spread is the artefact of an industry where the same product is sold by competitors that disagree, persistently, on how to price the same risk.

What the spread means for a borrower

The practical consequence is that "what can I borrow?" is a poorly-formed question. The well-formed version is "what's my panel-spread distribution?" — and the answer is a range, not a number, with the right lender for the file depending on which lever matters most for that file shape.

• A borrower with strong base income but no overtime sits well at lenders that don't aggressively shade non-base income — the assessment rate matters most for them.
• A borrower with material overtime or commission needs a lender with generous shading on those components — assessment rate is secondary.
• A self-employed borrower with volatile returns needs a lender with averaging methodology that suits the income shape — most-recent-year-capped vs lower-of-two-years can be a $200k swing on identical returns.
• A borrower at the DTI ceiling needs a lender that goes higher on DTI — assessment rate and shading don't matter if the cap blocks the file.

The broker's job, properly understood, is matching the file shape to the lever profile of the right lender — not chasing the cheapest rate, not optimising on a single dimension, not running a vanilla calculator and reading the answer back. The panel spread is the reason the broker job exists at all; if every lender priced identically, a calculator alone would do the work.

The honest interface

If borrowing capacity is a distribution, the honest interface for a borrower is one that returns the distribution. Most online calculators don't. The single-number model is operational shorthand — easier to render, easier for the borrower to grasp, easier for the lender to point at when something goes wrong. But it's misleading on the central question.

The site's borrowing capacity calculator returns the panel spread directly: same input, every active lender's policy applied, the spread visible as the result. The self-employed version does the same with averaging methodology and add-back rules surfaced per lender. The panel page shows the current state of the lever values across the 98 active lenders that produce the spread.

That's the honest version. Same file, different lender, different answer — and the answer matters. The broker file is the panel work.