AboutInsightsToolsSign in
All insightsRental markets
Oct 8, 202617 min

Why rent reports
disagree.
And what renters
actually pay.

Two national headlines. One lease. What changes when we look inside the numbers?

Zillow

+2.7%

September 2026 /
annual change

Released Oct 6

Apartment List

−0.4%

September 2026 /
annual change

Released Sep 29
Different estimates. Not interchangeable prices.
Source figures CSV
Look inside the numbers

The disagreement arrives before the lease.

Imagine a renter at the kitchen table with a renewal letter and two browser tabs. One tab says American rents are rising. The other says they are falling. The letter says the apartment will cost more next year. The renter has not yet made a mistake. Three different questions have arrived disguised as one.

This is a hypothetical scene. The opposing national figures are real. Zillow's October 6 report puts September's typical rent at $1,932, up 2.7% from a year earlier. Apartment List's September 29 report, headed October but measuring September, puts its national median at $1,388, down 0.4% over the year.

The temptation is to hold a quick trial. Somebody must have the wrong number. Or perhaps the lower estimate describes apartments and the higher one houses. Choose the explanation that suits the letter, close the other tab, and go back to worrying.

It is worth keeping both tabs open.

A rent report is a view through an instrument. Before asking whether the view is right, ask where the instrument is pointed. Is it watching asking prices? Likely prices at the moment a unit is leased? Payments made by people who moved years ago? Does its dollar figure describe the middle household, a carefully adjusted part of the available stock, or the lease in your hand?

The annual figures differ by 3.1 percentage points. That subtraction is easy. Explaining exactly how much each methodological choice contributes would require a common dataset and a controlled comparison we do not have. This article will not pretend that reading two methodology notes produces that experiment.

What those notes can give us is more useful than a verdict on which headline to believe: a way to ask a better question before making a costly decision.

01 / AdvertisedWhat is asked.

The price attached to the available home.

02 / ContractWhat is agreed.

The base rent and obligations in the lease.

03 / EffectiveWhat it averages.

Rent after credits, spread across the term.

Concepts, not three interchangeable national statistics. Fees and payment timing need separate attention.

The price of what, exactly?

There is no national apartment available for the national rent. The number has no keys, no commute, no second bedroom and no landlord willing to negotiate. It is a summary built from other people's homes.

The $544 separating the two published dollar estimates is therefore not a discount a renter has overlooked. It is not evidence that one website can find the same apartment for $544 less. The providers attach dollars to their estimates in different ways. Subtracting those levels and calling the result a saving gives a statistical difference the status of a shopping opportunity.

An index and a dollar benchmark do different work. An index tries to follow movement. A benchmark gives that movement a monetary reference. A bathroom scale could show that two people each lost 2% of their weight without suggesting that they should weigh the same. Likewise, a rent series can be useful for direction without its dollar level being an individual household's budget.

The distinction matters even within one report. A rate of change needs a comparison period. A price needs a unit of housing. A claim about affordability needs an income, an obligation and a household. Sliding between those questions is how a perfectly serviceable rent number becomes a misleading story.

For our renter, the next useful comparison is quite small: this apartment against genuinely available alternatives, with similar rooms, location, lease length and conditions. A national series can supply context. It cannot supply the missing apartment.

Nor can a falling national figure settle the fairness of a renewal increase. That argument needs local evidence and the actual agreement. A broad market trend may strengthen a negotiating position, but it is not the same thing as a comparable offer.

Inside the estimates / schematic, not a gap decomposition

A number is
a set of choices.

Both control for changing inventory. Their observations and dollar calibration are not identical.

Zillow / ZORI
  1. 01

    Asking rents

    Repeated observations of the same unit

  2. 02

    Stock weighting

    Structure size and building age

  3. 03

    Index + dollar reference

    Three-month smoothing; weighted middle 30% mean

Apartment List
  1. 01

    Likely leased units

    Last available rent before vacancy ends

  2. 02

    Same-unit comparison

    Repeated leasing episodes

  3. 03

    Index + Census reference

    Three-month smoothing; recent-mover adjustment

Public methods: Zillow and Apartment List. Last observed price is an inference, not a signed lease or proof that all concessions are captured. Diagram widths do not encode sample size, time or contribution to the 3.1-point gap.

The useful surprise in the methods is how much the providers have in common. Both try to avoid mistaking a change in the homes being observed for a change in their rents. Neither is simply averaging whatever luxury listings appeared this month.

Zillow's ZORI method follows asking-rent changes for the same units over time. Census-based weights account for structure size and building age. The series is smoothed over three months; its dollar reference uses a weighted mean of the middle 30% of recent asking rents, not a straightforward national median.

Apartment List's method also compares the same unit across leasing episodes. It infers a likely transaction price from the last available rent before a unit stops being vacant, then smooths the index over three months. A Census-based, recent-mover-adjusted rent benchmark supplies the dollar level. Its intended scope is broader than the units on its platform.

Those are materially different observations and calibration choices. But a unit disappearing from availability is not a signed lease we have read. And a last listed price does not, by itself, establish that every separate incentive is reflected. We should not promote an inference into more detailed contract evidence than the method provides.

Calling the comparison simply houses versus apartments would also be too neat. The providers' stated targets are not that clean split. Inventory coverage and weighting can matter without either index being a census of every rental agreement.

Good methods do not abolish disagreement. They make the disagreement more legible. To isolate it, researchers would need to hold the units, observation dates and price definitions constant, then change one choice at a time. Until then, the pipelines tell us what to investigate, not the percentage-point answer.

A market can look dearer without one landlord raising rent.

Consider a deliberately tiny, invented rental market. Ordinary units cost $1,600 a month. More expensive units cost $2,800. Those prices never change. In the first batch of twenty observations, ten are ordinary and ten are expensive. The average is $2,200.

Now imagine that sixteen of the twenty observations are expensive units. The average becomes $2,560. A careless chart could report a rent increase of more than 16%. Yet neither type of unit became dearer. The camera moved.

The experiment below lets you move it yourself. Each mark represents one hypothetical observation. Changing the mix changes the raw average while the two underlying prices remain fixed.

This is not a model of Zillow or Apartment List. It illustrates the problem that both providers' repeat-unit methods are designed to address. The point is not that their current disagreement can be blamed on this toy calculation. It is that an apparently obvious way of summarizing a market can answer the wrong question.

The same trap appears in ordinary conversation. A renter spends an evening seeing expensive new buildings and decides the city has become impossible. A landlord watches a few older units sit vacant and decides the city has collapsed. Both may be accurately describing their observations. Neither has yet established the citywide change.

Matching the same unit improves the comparison, but it leaves another question: which units are observed repeatedly? A method can control the mix inside its available evidence without making all evidence equally available. Missingness does not disappear because a regression is sophisticated.

That is a reason to read the design of an index, not a reason to dismiss indexes. The alternatives are often worse. What matters is whether the instrument was built for the question we are handing it.

An invented inventory / twenty observations

Move the camera.
Not the rents.

Every mark is one hypothetical observation. No unit changes price.

$1,600 ordinary unit$2,800 expensive unit
Raw sample average$2,200.00

Within-type rent change 0%

All ordinaryAll expensive
Illustrative arithmetic, not either provider's estimator. Both providers use repeat-unit methods to address composition changes. This exercise does not explain their current gap.

The empty apartment and the occupied apartment keep different time.

A vacant unit can be repriced tonight. A tenant halfway through a lease may still be paying the amount agreed months ago. These are not competing descriptions of the same payment. They are observations from different stages of the rental cycle.

In a hypothetical building with twelve identical leases, suppose everyone currently pays $1,800. A newly available unit would lease for $2,000. Replace one old lease with the new price and the building's average payment becomes $1,816.67. Replace all twelve and it becomes $2,000. The new offer did not rise twelve times. It spread through the building as agreements changed.

This is the mechanism in the lease clock below. The steps are imaginary leasing events, not actual months and not a forecast of CPI. No claim is made that twelve tenants renew on the same timetable or receive identical offers.

The distinction has a real research history. A 2022 BLS working paper by Brian Adams, Lara Loewenstein, Hugh Montag and Randal Verbrugge studied rent microdata. In their historical analysis, differences between new-tenant and all-tenant rent growth explained most of the disagreement among the indices they examined. New-tenant inflation led official BLS rent inflation by four quarters.

That finding is not a permanent twelve-month countdown. It is also not a decomposition of the current Zillow versus Apartment List gap: both of those private series look toward the market for available or newly leased housing. The study teaches a distinction, not a shortcut around today's evidence.

Official CPI rent measurement samples occupied rental housing, with units generally priced six months apart. Owners' equivalent rent uses rental evidence to measure housing services. It is not a mortgage-payment index or a claim that an owner literally sends a rent check to someone else.

There is another limit worth keeping in view. BLS says publication of its separate new-tenant and all-tenant research indices was paused in April 2026, following missing October 2025 survey data and competing priorities. That is not a suspension of the official CPI. Nor is it permission to invent current observations for the research series.

The lesson is to put a timestamp on the claim and identify whose rent it follows. A new mover, a continuing tenant and a national inflation statistic need not arrive at the same answer on the same day.

Twelve invented leases / a fixed cohort

One new offer.
Many old agreements.

The new rent stays at $2,000. Advance the leasing events to see the average catch up.

New offer$2,000.00

Average of twelve leases$1,816.67

01$2,000
02$1,800
03$1,800
04$1,800
05$1,800
06$1,800
07$1,800
08$1,800
09$1,800
10$1,800
11$1,800
12$1,800
1 of 12 repriced
A mechanism, not a calendar or CPI forecast. Each step replaces one $1,800 lease with a $2,000 lease. The cohort, prices and sequence are hypothetical.

The free month does not make the other months smaller.

Return to the kitchen table. On a third tab, a different building advertises one month free. For this hypothetical offer, assume $2,200 base rent, a twelve-month term, no other charges and the first month's rent waived.

Eleven payments of $2,200 make $24,200. Divided across twelve months, that is $2,016.67. The average is correct. So are the eleven larger payments. A household can understand the total cost perfectly and still run short in an ordinary payment month.

One number is a way to compare offers. The other is a cash obligation.

Spread exactly the same credit across all twelve payments instead, and each rent payment becomes $2,016.67. Total rent is unchanged. The household's payment calendar is not. A chart that preserves only the average cannot tell these two agreements apart.

This is why a rent report and a lease comparison should not be expected to contain the same information. Zillow reported a concession on 39.6% of its rental listings in September. That is a count of listings offering some concession. It does not tell us that 39.6% of renters received a free month, how large each benefit was, or whether every tenant qualified. Source: September market report.

An incentive might help a renter and still be an inconvenient way to help. If the benefit arrives late, a household must finance the months before it. If it arrives early, the renter needs to remember that the ordinary payment resumes. A genuine saving can have a demanding schedule.

The experiment lets you examine both. It does not assume an advertised promotion is unconditional. Eligibility, credit dates and any repayment clause have to be checked in the agreement. A calculator can add the amounts; it cannot supply a missing contract term.

Interactive / one lease, two kinds of cost

A free month.
A full-sized payment.

Change the term and charges. Then compare the average with the months you actually have to pay.

Contract base rent$2,200.00Before credits and fees

Effective monthly rent$2,016.67Total rent ÷ lease months

Average including modeled fees$2,016.67One-time fee spread for comparison

Payment calendar / USD per monthPeak payment $2,200.00Lease total $24,200.00
Rent after creditMonthly feeOne-time fee
The first rent payment is waived. Fees still apply. One-time fee is paid in month 1. Refundable deposits and other excluded costs are not shown.
No-credit renewal / same base rent9.1%

Change from initial effective rent to unchanged base rent, if the credit expires. Not a forecast of a base-rent increase.

Export CSVReturn to the result
See every payment and the calculation
Modeled payment schedule in USD
MonthBase rentCreditRent dueFeesTotal due
1$2,200.00$2,200.00$0.00$0.00$0.00
2$2,200.00$0.00$2,200.00$0.00$2,200.00
3$2,200.00$0.00$2,200.00$0.00$2,200.00
4$2,200.00$0.00$2,200.00$0.00$2,200.00
5$2,200.00$0.00$2,200.00$0.00$2,200.00
6$2,200.00$0.00$2,200.00$0.00$2,200.00
7$2,200.00$0.00$2,200.00$0.00$2,200.00
8$2,200.00$0.00$2,200.00$0.00$2,200.00
9$2,200.00$0.00$2,200.00$0.00$2,200.00
10$2,200.00$0.00$2,200.00$0.00$2,200.00
11$2,200.00$0.00$2,200.00$0.00$2,200.00
12$2,200.00$0.00$2,200.00$0.00$2,200.00

Total = base rent × (lease months − free months) + recurring fee × lease months + one-time fee. Whole free months only. The renewal comparison assumes unchanged base rent and no new credit.

An educational comparison, not a lease offer. Utilities, taxes, moving costs, refundable deposits and concession clawbacks are excluded. Check eligibility and the written terms of any real promotion. Calculated in your browser; CSV exports include the selected settings.

The smallest print can change the comparison.

Suppose another apartment has the same $2,200 rent and the same free month, but charges $75 every month and a $300 nonrefundable fee at the start. These are invented amounts, not a claim about a particular property.

The rent average is still $2,016.67. The charges add $1,200 across the year, making the average modeled cost $2,116.67. The promotion is real. So are the charges. Describing only one of them is a selective account of the same lease.

Now compare an otherwise equivalent offer at $2,100 with no free month and no modeled fees. Its twelve-month total is $25,200. The first offer's total is $25,400. In this example, the apartment with the more attractive promotion costs $200 more over the term.

Nothing here proves that the cheaper total is the better home. Perhaps the first apartment saves an hour of commuting each day. Perhaps its heating is more efficient. Perhaps it has a bedroom the household needs. The point of doing the arithmetic is to stop a promotional phrase from deciding a comparison that belongs to the whole household.

Refundable deposits need a separate line of thought. They can be a serious move-in cash requirement without being a permanent expense if fully returned. Treating the entire deposit as an annual cost would be misleading. Ignoring it in a cash plan would be misleading too. This experiment excludes refundable deposits rather than pretending those two questions are one.

Utilities, moving costs, taxes and the conditions of any concession are also outside the calculation. Add them to a real decision where applicable. The result shown here is the total of the included items, not a promise of total household housing expenditure.

For a renter, the practical document is a payment schedule: what is due, on which date, which part is refundable, and which advertised benefit depends on staying or meeting other conditions. A beautiful average should not hide an ugly month.

The rent can stay flat while the household's cost rises.

At the end of our first illustrative lease, suppose the landlord keeps base rent at $2,200 but offers no second free month. With no modeled fees in either year, average monthly rent moves from $2,016.67 to $2,200. That is an increase of about 9.1% in the household's average rent cost, with a 0% change in base rent.

The example is not a forecast of what landlords will do. It is what happens when a temporary benefit ends and everything else stays fixed.

A renter who paid eleven full months already knows the ordinary payment. A renter whose discount was spread across the year will see a larger recurring bill. The same total benefit creates two different experiences of its disappearance.

This is a small but important gap between the language of markets and the language of budgets. A series following base asking rents might describe stability while a household leaving an introductory arrangement describes a rise. Before accusing either account of being false, identify the amount that changed.

The landlord has a parallel problem. A lease with a credit may look stronger when reported at its face rent than when examined as collected rent over the term. A rental-property forecast should put the concession where the cash actually goes and treat renewal terms as assumptions, not as income already secured.

Our experiment shows a no-credit renewal comparison only to expose the mechanism. It holds base rent constant, assumes no new incentive and makes no claim about future market rents. If the initial rent is entirely waived, the percentage comparison has a zero denominator; the page says it cannot calculate that percentage rather than displaying an absurd infinity.

The right question for a household considering an introductory offer is not just whether the first year fits. It is whether the ordinary payment fits, and what alternatives exist when the introduction ends.

Back to the kitchen table

The country has
a rent statistic.
The household
has a due date.

Choose the question before choosing the number.

A renter shopping today needs comparable, available homes and their complete lease terms. A renter considering renewal needs the cost of staying against the cost and disruption of moving. A property owner estimating income needs expected collected rent, vacancy, incentives and expenses. An economist studying inflation needs a defined population and a consistent price concept.

These people can all find a national rent report useful. They cannot all use it in the same way.

For an individual search, start with location and usable housing: bedrooms, condition, accessibility, commute and the dates the household needs. Then compare the same lease horizon. A free month on eighteen months is a smaller proportional benefit than a free month on twelve. Converting both to a monthly average helps, provided the payment calendar remains beside it.

For a property forecast, test more than one renewal path. Keep unchanged base rent with an expired concession separate from a true increase in base rent. Keep an advertised unit separate from an occupied one. Do not turn a national annual change into next year's revenue growth for every home in a portfolio.

When evaluating the reports themselves, ask what each observation represents, which units enter more than once, how stock differences are handled, how dollars are attached to the index, and what changed since the last release. A moving average smooths noise, but a smooth line is not additional evidence about a particular apartment.

There may still be a question we cannot answer. We may not know which part of the current disagreement comes from sample coverage, observation timing, inferred transactions or calibration. Naming that limit does not empty the story of meaning. It keeps the useful mechanisms from being sold as a complete explanation.

Back at the table, both original tabs can stay open. The renewal letter is no longer being asked to prove which national report wins. The renter has a different task: find the actual alternatives, compare the full obligations, and work out which payment calendar the household can carry.

The country has a rent statistic. The household has a due date. Understanding the distance between them is where the decision begins.

Evidence, definitions and the experiment.

Evidence checked October 8, 2026. The national figures refer to September, despite different release dates and the October heading on Apartment List's report. These are a dated editorial snapshot, not a live data feed. Future releases may revise estimates.

The two national dollar benchmarks are not directly interchangeable offers. The method diagrams are schematics, not measured sample sizes or a decomposition of the annual growth gap. The inventory and lease-clock exercises are deterministic hypothetical examples, not replicas of provider estimators or projections of CPI.

The lease experiment computes base rent × (lease months − free months), adds recurring fees × lease months and the one-time fee, then divides by lease months for its average. Whole free months are modeled. Upfront credits waive the first specified rent payments; spread credits are allocated equally across the term. The one-time fee is charged in month one. Charges and credit eligibility in an actual lease may differ. No interest, tax treatment, refundable deposits, utilities, moving expenses or concession clawbacks are modeled.

Displayed dollars round to cents. CSV payment rows retain six decimal places; tiny rounding differences can arise when separately rounding each payment. No inputs are sent to a lender, landlord or database.

Primary sources:

Read as Markdown Explore lease renewal
More insights