Skip to content

When the Mortgage Machine Prefers to Be Wrong

By Peter Yoder. . 5 minute read.

Why the Most Affordable Homes in America Almost Never Sell

A duplex, triplex, or fourplex should be the most forgiving path into homeownership. The rent from the other units cushions the mortgage, diversifies the borrower’s risk, and lowers the monthly net cost below renting.

Yet applications for 2–4 unit properties are denied at nearly twice the rate of single-family homes, and turnover in this segment runs roughly four times lower than the detached single-family average.

We built a single-family manufacturing line and then asked it to process an income property without giving the line new tools, new checkpoints, or new incentives. When the workflow is mismatched to the asset, the safest move inside the machine is to say no. And so it does.

Underwriters are punished for being wrong and rarely rewarded for being right. On a one-unit house, the system has decades of muscle memory: one appraisal form, a clear income calculation, an automated engine that recognizes most patterns, and a capital market that knows how to price the risk.

On a duplex, triplex, or fourplex, the file becomes two things at once: a home and a small business with rents. The forms multiply, the math becomes manual, and the gray areas grow. In a world where errors trigger buybacks, insurance rescissions, or examiner attention, the rational response to gray is to decline.

In 1936, with the FHA Charter, the United States did something remarkable: it allowed people to buy up to four units with the same fixed-rate, long-amortization mortgage used for a single-family home.

That decision recognized a truth about household finance: future rent can subsidize present ownership. It also created a classification problem. A fourplex is not a 200-unit apartment building with audited financials, but it is not a detached home either. It sits between categories: too small for multifamily underwriting, too dependent on rental evidence for the single-family factory.

Institutions solve category problems with paperwork. We added rent schedules, small-income property appraisals, and worksheets. We did not rebuild the line. We bolted complexity onto a process built for simplicity.

A single-family loan asks: does this household earn enough to carry this debt, and is the home worth what we think?

A small multifamily loan asks that plus: what is the market rent for the other units, are those rents real and durable, what vacancy should we assume?

None of those questions is unreasonable. The problem is that each one introduces manual judgment in a system tuned to automation. An underwriter must reconcile leases with an appraiser’s opinion of market rent. Someone must apply a vacancy haircut and decide whether to use the lease figure or the market figure. Every additional step is a surface where defects can appear later in quality control. Every surface invites a conservative overlay: “show me more proof,” “tighten the assumption,” “raise the reserves.”

If you hold the pen and the career risk, you do what rational people do under one-sided loss: you require perfection or you decline.

Credit risk in modern housing finance is shared among the borrower, the lender, the insurer, and the investor. Process risk is not. Process risk lives with the lender. If the file is missing a lease page, if the rent analysis contradicts the appraisal, if the vacancy factor is misapplied, the investor can force a repurchase or the insurer can rescind coverage.

The cheapest path for a lender is to avoid ambiguous files. The decision maker carries the downside of a mistake; the upside of a good approval accrues to the borrower and to the abstract cause of housing access.

Here the convexity is negative: small documentation errors have outsized costs. In that landscape, the optimal individual strategy is fragility-avoidance: short volatility, not access maximization. The institution is doing what it was paid to do. The social outcome, however, is that the safest borrowers in substance (steady job, one vacant unit for them to occupy, three others paying rent) look riskier in process.

Rent is not treated as data, it’s treated as paperwork. Leases sit in PDFs, appraisers write narratives, bank statements hide payment history. Nothing is machine-readable, so every file is an exception. Vacancies are guessed at, not standardized, so one appraiser’s optimism collides with another lender’s haircut.

The math that decides whether a borrower qualifies is done manually, precisely at the point where underwriters know they will be audited later. Time works against the borrower: while conditions pile up, rate locks expire and patience runs out. Each lender adds its own “safety checks.” Individually minor, together they become a wall.

Markets ration in three ways: price, quantity, or friction. Price is capped by regulation, supply is constrained by zoning and policy. What remains is friction. 2-4 unit loans are not scarce because the borrowers are unqualified; they are scarce because the process makes them hard to close. Denial rates climb, contracts collapse, turnover slows.

The hidden tax is paid by renters who could own but are pushed back to the sidelines. The neighborhood loses a resident owner and gains another absentee landlord. The system congratulates itself for being “prudent.”

We conflate prudence with purity. Prudence is protecting against real loss. Purity is protecting against blame. Institutions choose purity because defect-free files are rewarded, even if worthy borrowers are excluded. A denial is invisible; a repurchase is career-ending. The outcome is predictable: we polish the ladder’s rungs instead of letting people climb.

A mortgage system can optimize for clean files or for accurate inclusion. For small multifamily, we have chosen the first at the expense of the second. The result: higher denial rates, longer cycle times, and the erosion of the best on-ramp to ownership.

The way out isn’t new subsidies or new ideology, just a workflow that recognizes what it is processing. Rent should be treated as structured input, not as a liability. When that happens, overlays retreat, approvals rise, and the system begins doing what it was designed to do: match credit to reality.

There will be no ribbon-cuttings for an automation that reconciles a lease to an appraisal. But someone will move into the vacant unit on the top floor, and three neighbors’ rent will help them make the payment, and a block that was drifting toward absentee ownership will gain a resident owner with skin in the game. That is what finance is for: not to defend a process from embarrassment, but to match reality with credit so people can build durable lives.