Two Housing Reports Just Dropped. Both Say Prices Moved 2.5%. One Means a Crash, the Other a Record High.

Two housing reports dropped 24 hours apart this week. Both said "2.5%." One meant a historic crash. The other meant a record high.

On July 1, Realtor.com told the world that home prices just posted their steepest annual decline since the report started tracking data in 2017. On July 2, Redfin said home prices just hit an all-time high. Nobody fat-fingered a spreadsheet. Both numbers are correct — and investors who can't explain why are one bad comp away from mispricing an offer or overpaying for a rate lock.

This isn't a one-off fluke. It's what happens when a market sits at the exact seam between a buyer's market and a seller's market: every data provider is measuring a slightly different slice of the same transaction funnel, and right now those slices are pointing in opposite directions. The gap between them is where deals get won or lost — not in the headline number, but in knowing which number applies to the property in front of you.

The same split shows up in mortgage rates. On the same two days these price reports came out, four legitimate, widely cited sources quoted "the 30-year rate" anywhere from 6.36% to 6.73% — a 37 basis point spread on a number most investors treat as a single fact. That's real money on a rate lock, and it's avoidable if you know what each source is actually measuring.

Ask vs. Sold: The Metric Gap That's Eating Headlines Alive

Realtor.com's number is an asking price — pulled from active MLS listings, a leading indicator of what sellers think their home is worth right now. Redfin's number is a closed sale price — a trailing indicator of what buyers actually paid on deals that finished escrow in the last four weeks. Those are two different stages of the same funnel, and in a market like this one, they can move in opposite directions without contradicting each other at all.

Here's the mechanism. Realtor.com's chief economist, Danielle Hale, put it plainly in the June report: sellers are "pricing accordingly from the start rather than listing high and cutting later." That behavioral shift pulls the asking number down immediately — sellers are front-loading the discount instead of chasing the market down over months of price cuts. Meanwhile, the homes that actually close are a different, smaller population: better-located, already-adjusted, or simply the ones motivated enough to get to the finish line. That mix skews the closed number up even while list prices are falling. It's not a contradiction. It's two different measurements of two different populations of homes, seven days apart.

Layer in geography and it gets sharper. Realtor.com's senior economist Jake Krimmel calls it "the two Americas story" — now four years running. The West is down 4.0% year-over-year and the South is down 2.5%, both still working through affordability limits. The Northeast is down just 1.0%, and the Midwest is flat, both held up by tight supply. At the metro level the spread is even wider: Austin (-8.2%), Memphis (-6.0%), and Buffalo (-5.2%) on the losing end; Providence (+8.7%), Indianapolis (+4.9%), and New York (+3.4%) on the winning end. A national "average" blending all of that tells you almost nothing useful about the deal on your desk.

The Numbers

Realtor.com June 2026 report (published July 1):

• Median asking price: $430,000, down 2.5% year-over-year — the steepest annual decline since the report began in 2017
• Pending sales: up 3.7% year-over-year, the seventh straight month of gains
• Median days on market: 53 days, flat year-over-year — the first non-increase in 26 months
• Regional price change (YoY): West -4.0%, South -2.5%, Northeast -1.0%, Midwest flat
• Weakest metros by price-per-square-foot: Austin -8.2%, Memphis -6.0%, Buffalo -5.2%
• Strongest metros: Providence +8.7%, Indianapolis +4.9%, New York +3.4%

Redfin four-week tracker ending June 28 (published July 2):

• Median sale price: $408,838, up 2.5% year-over-year — a record high
• Median monthly housing payment: $2,633, up 1.4% year-over-year — the first YoY increase since October
• Pending sales: up 0.4% week-over-week, up 2% year-over-year
• New listings: up 1.1% week-over-week, up 1.7% year-over-year

The mortgage rate spread, all quoted within the same 48-hour window:

• Freddie Mac PMMS (weekly survey): 6.43% on July 2, down from 6.49% the prior week — a seven-week low
• Redfin daily average: 6.49% for the week ending June 25
• HousingWire's Mortgage Rates Center (Polly locked-rate index, actual closed loans across all credit tiers): 6.73%
• Norada Real Estate daily rate roundup: 6.36% on July 2

That's a 37-basis-point gap on "the" 30-year rate, reported the same week. On a $320,000 loan (80% LTV on a $400K purchase), the difference between 6.36% and 6.73% is roughly $75 a month — about $900 a year, and close to $27,000 over the life of the loan if it's baked into your rate assumption and never questioned.

Common Mistakes Investors Make Here

• Quoting a national median in an offer or BPO without checking whether it's asking or sold. A seller anchored to a "record high" headline is reading Redfin's closed-sale number; your comp should be built from actual closings in their zip code, not the national print.
• Treating one 4-week window as a trend instead of noise. A short tracking window can swing hard on mix-shift alone — which specific homes happened to close — especially in thinner markets.
• Locking a rate off a marketing-site "as low as" quote instead of an actual locked rate for your credit profile and property type. Investment-property and DSCR rates run above every owner-occupied number you'll see in a headline.
• Applying the national trend to a local deal. Austin and Providence moved 17 points apart in the same report. A national average is close to useless for underwriting a specific address.

How to Use PropGPT for This

"Pull the last 90 days of closed sale prices vs. current active listing prices for [ZIP CODE], and show me the average discount-to-list percentage."

This tells you whether your target market is behaving more like Realtor.com's asking-price story or Redfin's closed-price story, instead of guessing from a national headline.

"Compare the year-over-year price trend for [METRO] against these national numbers — Realtor.com asking prices -2.5% YoY, Redfin sale prices +2.5% YoY — and flag whether this metro is tracking the national trend or diverging from it."

Feed PropGPT the figures above and get an instant read on whether your market is behaving like a Providence (still climbing) or an Austin (still falling).

"Model my monthly payment on a $[PURCHASE PRICE] investment property at 6.36%, 6.43%, 6.49%, and 6.73% — show me the dollar difference at each rate over 12 months and over the full loan term."

Turns the rate-divergence confusion into a hard number, so you know exactly what a lender's quote is costing you relative to the best rate reported that week.

"Show me the days-on-market trend for [ZIP CODE] over the last 6 months and tell me if it's rising, flat, or falling faster than the national 53-day median."

Days-on-market flattening nationally doesn't mean it flattened in your market. This confirms the local picture before you adjust your offer strategy.

"Given a seller in [ZIP CODE] anchored to a 'record high' price headline, draft a comp-based counter-offer using actual closed sales from the last 90 days, not asking prices."

Turns the ask-vs-sold gap into an actual negotiating script for a live deal, instead of an abstract data point.

The Bottom Line

Two reports, one week, both technically right — because "the housing market" isn't one number. It's a dozen local, stage-specific numbers, and right now the gaps between them are wide enough to swing a deal. The investors making money in this exact moment aren't the ones repeating whichever headline they saw first; they're the ones pulling their own zip-code-level ask-vs-sold comps and their own actual rate quote before they write an offer or sign a lock.

A lazy national price comp or a lazy 37-basis-point rate assumption can eat your entire margin on a deal this size. Before your next offer, run the ask-vs-sold pull for your target zip code in PropGPT — it takes five minutes, and it beats finding out the hard way which "2.5%" you were actually negotiating against.