Multifamily Bad Debt Underwriting: Beyond the 1% Plug

Published October 6, 2026

Bad debt write-offs averaged $4.2M per respondent over a 12-month period in the National Multifamily Housing Council (NMHC) Pulse Survey fielded between November 2023 and January 2024, against a median of roughly $800,000 across the same 75 apartment owners, developers, and managers. That distance between the mean and the median is the whole problem with a flat bad debt assumption. The line is not distributed evenly across portfolios, and it is not distributed evenly across jurisdictions — yet most acquisition models still carry it as one percentage of Gross Potential Rent (GPR), frequently 0.5% or 1.0%, applied with equal confidence to a 250-unit asset in metro Atlanta and a 250-unit asset in suburban Minneapolis.

Why Multifamily Bad Debt Underwriting Fails at a Flat 1%

Bad debt is not an input. It is the residual of three separate processes: who was approved at the door, who stopped paying once in place, and how long the jurisdiction takes to return the unit. A single percentage models none of the three, so it cannot respond when any of them changes.

The direction of travel is also not flat. Per the National Apartment Association's (NAA) survey of operating results, bad debt rose about 23% year over year from 2021 to 2022, a period in which gross rent growth was still running near double digits — so the loss line was growing faster than the revenue it was deducted from. Multifamily bad debt underwriting that carries a constant percentage through a five-year hold implicitly assumes the opposite, that collection loss scales neatly with billed rent. Operating results vary widely by market and operator, and a single national survey year is not a trend.

The cleanest way to think about it is that bad debt belongs on its own schedule, in the same way the property tax line deserves an assessed value and a millage rate rather than a trended figure from the seller's statement.

The T-12 Often Nets Collection Loss Out of Income

The first obstacle is that the number is frequently invisible on the document a buyer is handed. Many property-management exports book write-offs as contra-revenue inside rental income, or against the month a resident skips, so a trailing-12 statement can present net collected rent with no separate loss line at all. A model built on that statement is not assuming 1.0% bad debt — it is assuming whatever the seller's accountant already netted out, which is generally unknowable from the summary.

Reading a T-12 statement for collection loss usually means asking for different documents: the delinquency aging report as of the cutoff date, the write-off register by month for 24 months, and the resident ledger detail for any unit with a balance over about 30 days. Those three reconcile to a real loss rate. The summary statement generally does not, which is why collection loss sits alongside taxes and insurance among the lines that have to be rebuilt rather than trended.

One detail is easy to miss. A seller preparing an asset for market has an incentive to write off aged balances early or late depending on which presents better, and either choice distorts the trailing rate. Comparing the write-off register against the aging report across the same months is roughly how that gets caught.

Eviction Velocity Is the Variable, and It Is Local

The time between a missed payment and a returned unit is approximately the window over which billed rent accrues with little realistic prospect of collection. That window is set by county practice, not by the operator.

The dispersion is wide. Per the Eviction Lab's preliminary analysis of 2025 filing patterns, landlords filed about 1.23 million eviction cases across tracked sites last year, roughly 3% below the post-pandemic average, at an average filing rate near 8% — approximately one case for every 13 renter households. Within that average, the Eviction Lab reports Atlanta at the high end with close to one case per four renter households, New York City near 5%, and Travis County, Texas running about 30% above its own baseline. Filing rates are a measure of process volume rather than of dollars lost, and the tracked sites are not a national sample.

Timelines compound the same geography. Georgia's dispossessory process generally runs one to three months depending on contestation and court calendars, per iPropertyManagement's 2026 summary of the state's procedure, which also notes added delay in high-volume counties such as Fulton and DeKalb. Each additional month of possession lag is, very roughly, one more month of billed rent at risk per nonpaying unit, plus the make-ready that follows. Court practice changes, so current local timelines are worth verifying with counsel in the actual venue rather than inferred from a state-level summary.

Application Fraud Sits Upstream of the Bad Debt Line

Screening quality is the other half of the equation, and survey evidence suggests it has been under pressure. In the same NMHC Pulse Survey, over 90% of respondents reported experiencing fraud in the prior 12 months and over 80% reported falsified pay stubs, employment references, or income documentation. About a quarter of respondents' bad debt was attributed, on average, to nonpayment on fraudulent applications, and a similar share of eviction filings traced to the same source. Nearly half of those observing variation by jurisdiction named Atlanta specifically.

This matters for underwriting because it breaks the usual assumption that a trailing loss rate is a property characteristic. If a seller's in-place roster was approved on documentation that would not clear a tighter screen, the embedded loss travels with the rent roll into the buyer's first year, and it is not visible in a cap rate comp. That is the kind of exposure a red-flag review of the rent roll is meant to surface before it reaches the model.

The survey covered 75 self-selected NMHC and NAA member firms over a single period, so these figures are generally indicative of direction and relative severity rather than a measured national fraud rate.

What Collection Loss Costs: A Worked Example

Example — illustrative. A 250-unit Class B property in a Sun Belt secondary market. Average in-place rent of $1,450 per month, so GPR of about $4.35M. Vacancy at 7%, concessions at 2% of GPR, other income at 5% of GPR, and operating expenses of $2.32M — roughly $9,280 per unit. The only variable is bad debt: 1.0% of GPR in the left column, 2.5% in the right. Debt is sized at 60% Loan-to-Value (LTV) on the left-column value, $18.9M at 6.0% with 30-year amortization, for annual debt service near $1.36M.

The spread is 150 basis points of GPR, or about $65,000 a year — roughly $261 per unit. Capitalized at the going-in rate it is close to $1.1M of value on a $31.5M asset, which is generally wider than the final negotiating gap on a deal that size. On coverage, DSCR slips from about 1.33x to about 1.28x, crossing below the 1.30x threshold some lenders carry and most sponsors watch.

Rents, expense loads, and cap rates vary substantially by submarket and vintage, so the figures above are illustrative rather than indicative of any particular market. The structural point is the one that travels: a loss line small enough to look like a rounding convention moves NOI by about 3.6% in this example, and the gap between a 1.0% convention and a 2.5% reality is not an unusual spread between jurisdictions.

Where This Analysis Breaks

The strongest counter-argument is that bad debt is partly controllable, which makes the market figure the wrong anchor for a capable operator. A sponsor with tighter income verification, same-week filing discipline, and a functioning collections process can genuinely run below a submarket average, and underwriting the average could overstate loss and cost that sponsor deals it should win. The honest version is to underwrite the market rate as a base case and treat the operating improvement as upside, not to assume it in year one.

Double counting is the opposite error and it is easy to commit. If the seller's statement already nets write-offs out of rental income, layering a fresh 2.5% bad debt line on top deducts the same loss twice and manufactures a deal-killer that is not there. The reconciliation between the aging report and the write-off register is what distinguishes the two cases, and skipping it makes either direction of error equally likely.

Filing rates are also not loss rates. A jurisdiction with a fast, inexpensive process may show a high filing rate precisely because filing is used as a routine collections tool, and that can coexist with lower ultimate write-offs than a slow venue where operators file rarely and absorb longer arrears instead. Atlanta's elevated rate in the Eviction Lab data is consistent with either reading, and the data alone does not settle it.

Finally, the NMHC dollar figures describe whole portfolios per respondent, not per unit or per asset, so they cannot be scaled down to a single property. They establish that the line is large and unevenly distributed. They do not establish what it is on any specific deal.

The Underwriting Takeaway

Three things follow. Bad debt deserves its own line derived from the subject asset's write-off register and delinquency aging, rather than a house percentage inherited from the template. The assumption belongs next to the venue's actual possession timeline, because the loss is largely a function of how long a nonpaying unit stays occupied. And the reconciliation against the seller's netting convention comes first, since a 2.5% assumption on top of an already-netted statement is as wrong as a 1.0% assumption on a gross one — the same economic-occupancy discipline that a revenue model built on occupancy rather than asking rent demands everywhere else.

A unit that is occupied, billed, and not paying is the most expensive kind of vacancy, because the model counts it as income.

UWmatic is an AI-powered underwriting platform built for multifamily and mobile home park investors. Model collection loss as its own line against the delinquency aging and see what a 150 basis point swing does to NOI, DSCR, and value. Try the free underwriting calculator →

This analysis reflects current market interpretations as of the publication date and may evolve as new data becomes available. Figures cited are drawn from public sources including the National Multifamily Housing Council Pulse Survey on rental application fraud and bad debt, National Apartment Association survey of operating results, the Eviction Lab's preliminary analysis of 2025 eviction filing patterns, and iPropertyManagement's 2026 summary of Georgia's eviction process, and are subject to revision. Nothing in this post is investment advice; readers should conduct their own diligence and consult qualified professionals before making investment decisions.