The Mortgage Bankers Association's May 20 Weekly Applications Survey told a quiet story under a noisy headline. The Market Composite Index slipped 2.3% for the week ending May 15, and the purchase index fell 4%. Underneath those numbers, something else moved: the adjustable-rate mortgage share of applications climbed to 9.6% — the highest reading since October 2025 — while the MBA's 30-year fixed contract rate sat at 6.56% and the 5/1 ARM contract rate came in at 5.76%. Two days later, Freddie Mac's Primary Mortgage Market Survey pegged the 30-year fixed at 6.51%, up from 6.36% the prior week.
An 80-basis-point gap between the 5/1 ARM and the 30-year fixed is real money. It is also the smallest, most negotiable-sounding number on a Loan Estimate — until the fifth anniversary of the loan, when the cap structure does what cap structures do. For Georgia buyers staring at $400K-and-up purchase prices in the Atlanta metro, the ARM question deserves a worked example, not a vibe.
The spread, in plain dollars
Set the scene with a number that maps to actual Georgia inventory. The FRED median list price for the Atlanta-Sandy Springs-Roswell metro was roughly $422,400 in April 2026; cross-checks against Redfin's Atlanta market page put sold-price medians in a similar band. A $415,000 purchase price with 20% down — a $332,000 loan — sits squarely at the metro midpoint.
At the MBA's contract rates for the week ending May 15:
- 30-year fixed at 6.56%: approximately $2,112/month principal and interest.
- 5/1 ARM at 5.76%: approximately $1,940/month principal and interest.
The monthly delta is about $172. Held flat through the entire 60-month fixed window of a 5/6 ARM, the cumulative principal-and-interest savings clear $10,000 before the first reset is even on the calendar. A 7/6 ARM typically prices slightly above a 5/6, but it buys two extra years of payment certainty — which is the variable that actually matters for most buyers, not the headline rate.
How the reset actually works
The reason ARMs are a tool, not a trick, is that the post-reset payment is not arbitrary. It is mechanical, and the inputs are public.
Conforming 5/6, 7/6, and 10/6 ARMs almost universally reference the 30-day average SOFR published by the New York Fed. As of May 18, 2026, that index sat at 3.625%. At reset, the lender adds a fixed margin — commonly between 2.75% and 3.00% on conforming product, per Pennymac's correspondent SOFR ARM overview — to arrive at the fully indexed rate. With today's index and a 2.875% margin, the math points to roughly 6.50% if a reset happened tomorrow. That's not the post-reset rate, however; the rate cap is.
The CFPB's plain-English explainer walks through the three-number cap structure that has to appear on the Loan Estimate: initial adjustment cap, subsequent (periodic) adjustment cap, and lifetime cap. Two structures dominate conforming product:
- 5/6 ARMs: typically 2/1/5. The rate can move up to 2 percentage points at the first reset, 1 point at any subsequent six-month adjustment, and 5 points above the start rate over the life of the loan.
- 7/6 and 10/6 ARMs: typically 5/1/5. The first reset can jump up to 5 points immediately, then 1 point per period, with the same 5-point lifetime ceiling.
Apply that to the example loan. A 5/6 ARM starting at 5.76% with a 2-point initial cap can rise to 7.76% at month 61. On the $332,000 loan, the balance after five years of amortization at 5.76% is roughly $307,000, and a 25-year reamortization at 7.76% lands the new payment in the mid-$2,300s — a payment jump on the order of $380/month from the starting figure. That's the cap doing its job; the actual reset rate would equal the fully indexed rate or the capped rate, whichever is lower. The 7/6 structure looks gentler on the surface but is harsher in the worst case: a 5-point first-reset cap means a 5.76% start rate can become 10.76% at month 85.
This isn't theoretical. The CFPB's Consumer Handbook on Adjustable-Rate Mortgages — the booklet lenders are required to provide — includes an Adjustable Interest Rate (AIR) table that spells out the maximum possible payment under the contract's caps. The number is on the document. Borrowers don't read it.
Where the loan fits in Georgia
An ARM fits when the borrower's exit predates the reset. In Georgia, two profiles fit cleanly.
The PCS rotation. Georgia hosts Fort Stewart and Hunter Army Airfield (3rd Infantry Division), Fort Moore (the Infantry and Armor schools), Robins AFB, and Moody AFB. Permanent change of station orders for active-duty personnel typically run on 2-to-4-year cycles; the Fort Stewart inbound PCS page documents the cadence. A 5/6 ARM on a Hinesville-area purchase — where medians sit in the $209K–$270K band, comfortably under the 2026 Fort Stewart E-5-with-dependents BAH of $2,310/month — will almost always outlast the assignment. The 60-month fixed window matches the operational reality.
One caveat for the same buyer pool: VA loans are assumable. If a target property has a sub-4% VA loan originated in 2020 or 2021, an assumption can beat any new origination — ARM or fixed — by a wide margin. The ARM question only matters if the seller's existing financing isn't on the table.
The intentional 5-to-7-year intown hold. Buyers playing transitioning Atlanta submarkets — parts of Edgewood, Pittsburgh, Sylvan Hills, the western half of West End — often have a stated thesis: hold 5 to 7 years, capture appreciation, exit. If the exit is on the calendar before the reset is, the ARM is the appropriate instrument for the trade. The fixed-rate premium is insurance against a contingency the borrower has already ruled out.
Where it's a slow-motion mistake
The same 80-basis-point spread that funds the right buyer's strategy underwrites the wrong buyer's payment cliff. Two profiles to flag.
The DTI-stretched first-time buyer. If the borrower only qualifies on the lower ARM payment — if the 30-year fixed pushes the debt-to-income ratio over the lender's threshold — the cushion that lets the household absorb a reset doesn't exist. Refinance optionality exists in theory, but it is a function of rates at month 61, not rates today. Counting on a refi to bail out a stretched DTI is counting on the rate environment to cooperate on a specific calendar date five years from now.
The retiree in the forever home. An ARM on a fixed income, with no planned exit and no income trajectory to absorb a reset, is a structural mismatch. The 2-point first-reset cap on a 5/6 — or worse, the 5-point cap on a 7/6 — becomes a budget event with no lever to pull. For this buyer, the 80 bps of monthly savings buys nothing that matters and exposes everything that does.
The Loan Estimate checklist
Before signing on any ARM, the Loan Estimate should answer six questions:
- What are the initial, periodic, and lifetime adjustment caps? (The CFPB requires these to be disclosed; verify the numbers, don't infer them from the product name.)
- What index does the loan reference, and what is the margin? (30-day average SOFR + lender margin is the typical conforming structure.)
- What is the first adjustment date, to the month?
- What does the AIR table show as the maximum possible monthly payment?
- Is there a prepayment penalty? (Rare on conforming product, but verify.)
- For VA-eligible buyers: has the search included assumable VA loans on the target list? An assumption of a sub-5% VA note can dominate any new ARM or fixed origination.
The macro caveat
Bright MLS chief economist Lisa Sturtevant, quoted in a May 21 Real Estate News piece, described the late-spring outlook as "darkening" with rates above 6.5% and no easing signals on the immediate horizon. Realtor.com senior economist Anthony Smith, in the same piece, attributed recent rate moves primarily to Middle East geopolitics rather than domestic policy — a driver that's difficult to time and easier to underestimate.
That backdrop matters for one specific reason: the 80-bp ARM-vs.-fixed spread today is narrower than the 150-to-200-bp spreads that historically made the ARM trade-off obvious. Cross-checks against retail product on the article's publication date — NerdWallet's 5-year ARM comparison and Yahoo Finance's same-day rate snapshot — confirm the headline gap but not always with the same offer rates a specific borrower will see. The Loan Estimate, again, is the document that matters.
Bottom line
ARMs aren't the villain of the late-spring 2026 mortgage market, and the borrowers nudging the ARM share to 9.6% aren't all making a mistake. The 80-basis-point spread is real money for the buyer whose exit predates the reset — the PCS rotation, the intentional intown 5-to-7-year hold — and a payment cliff for the buyer whose exit doesn't exist. The math doesn't care which one you are. The Loan Estimate spells out both outcomes. Read the AIR table.
Related reading
Sources
- Mortgage Bankers Association — Mortgage Applications Decrease in Latest MBA Weekly Survey (May 20, 2026)
- Fox Business — Mortgage rates rise to 6.51%: Freddie Mac (May 21, 2026)
- Real Estate News — Rates up, sales down as housing market tug of war continues (May 21, 2026)
- FRED, St. Louis Fed — 30-Day Average SOFR (SOFR30DAYAVG)
- Federal Reserve Bank of New York — SOFR Averages and Index Data
- CFPB — What are rate caps with an adjustable-rate mortgage (ARM)?
- CFPB — Consumer Handbook on Adjustable-Rate Mortgages (CHARM Booklet)
- Pennymac Correspondent — Conforming SOFR ARM Overview
- FRED, St. Louis Fed — Median Listing Price, Atlanta-Sandy Springs-Roswell, GA
- Redfin — Atlanta Housing Market
- Veteran.com — 2026 Georgia BAH Rates
- U.S. Army Garrison Fort Stewart — Inbound PCS
- NerdWallet — Compare 5-year ARM Mortgage Rates (May 22, 2026)
- Yahoo Finance — Mortgage and refinance interest rates today, May 22, 2026
