The 2024 housing market: what the primary sources actually showed

WHAT YOU’LL LEARN

Reconstruct 2024 with explicitly dated primary-source releases
Separate national price appreciation from local acquisition evidence
Translate rate movements and construction signals into an auditable property model

Research coverage: 2024. Publication date is shown separately above.

Read 2024 through an evidence ledger, not a remembered headline

A useful history of the 2024 housing market has to answer two different questions: what happened during the year, and what an investor could reasonably have known before making a decision. This reconstruction uses official releases covering 2024, including publications issued in early 2025. It is written with a research cutoff of October 27, 2026. That later research date permits checking the record; it does not turn later revisions into information that existed when a 2024 purchase contract was signed. The examples below are invented teaching cases, not academy transactions or reported investor results.

The working conclusion is an interpretation: 2024 rewarded separating financing, price direction, construction supply, and individual property economics instead of forcing them into one national story. A rising index did not establish that a particular renovation would earn its target margin. A temporary improvement in a mortgage benchmark did not guarantee financing at the eventual closing. To make this history usable, each factual observation below names its period and release vintage, while each underwriting conclusion remains an explicit application or inference rather than a finding claimed by the statistical agency.

The price record: positive nationally, uneven underneath

FHFA's February 25, 2025 release reported a 4.5 percent increase in its purchase-only house price index between the fourth quarters of 2023 and 2024. The fourth-quarter increase from the previous quarter was 1.4 percent. In that release, 49 states posted annual increases; Mississippi declined 0.2 percent. The Middle Atlantic division gained 7.1 percent over four quarters, while West South Central gained 2.3 percent. These are nominal index changes for specified periods, not a measure of inflation-adjusted investor profit or the appreciation of every house.

The same release showed why the national figure could mislead an acquisition file: 92 of the 100 largest metropolitan areas rose over four quarters, while Cape Coral-Fort Myers declined 6.3 percent. An investor researching a local house should therefore preserve the metropolitan result as context and then examine nearby completed transactions, condition, concessions, and competitive listings. The original inference here is modest: a national gain can coexist with substantial local downside. Using the national number as an automatic adjustment to every comparable sale would erase precisely the variation the regional evidence reveals.

A mortgage-rate path with more than one turning point

Freddie Mac's archived 30-year fixed-rate PMMS observations were 7.22 percent on May 2, 2024, 6.08 percent on September 26, and 6.85 percent on December 26. These three dated observations demonstrate a decline followed by a rebound; they are not a claim about every weekly movement or a quotation available to a specific investor. The archive is the appropriate place to check the exact week associated with a financing assumption. Selecting only September's lower observation would produce an incomplete account of the year.

Freddie Mac's methodological note describes an application-based benchmark selected for conventional, conforming, owner-occupied purchase lending with particular credit and leverage characteristics. That makes PMMS useful historical context, not a direct substitute for a rental-property, bridge, rehabilitation, or commercial loan proposal. In a reconstructed investment memo, record the investor's actual lender terms separately. Include amortization, points, interest reserves, maturity, and any conditions preventing funding. The historical benchmark then explains the broader rate environment while the documented proposal determines the cash obligations the property actually has to support.

New-home sales: preserve the January 2025 vintage

The Census Bureau and HUD's January 27, 2025 release estimated 683,000 new homes sold during calendar 2024, compared with 666,000 in 2023. It also reported December sales at a seasonally adjusted annual rate of 698,000, an end-December inventory of 494,000 new houses for sale, and 8.5 months of supply. The estimated 2.5 percent annual sales increase carried a 90 percent confidence interval of plus or minus 4.5 percentage points. Because that interval includes zero, describing the change as a statistically established increase would overstate that release.

This is a release-vintage account, not a claim that these estimates were never revised. The December 2025 sales release, issued February 20, 2026, used a revised 2024 annual sales figure of 686,000. Both numbers belong in a careful research file with their publication dates. The first answers what the early year-end report said; the second supports comparisons made within the later release. Mixing an unrevised numerator from one vintage with a revised denominator from another can manufacture a growth rate that neither source actually published.

Construction showed a pipeline, not just one supply number

The January 17, 2025 Census-HUD construction release estimated 1,364,100 housing starts for calendar 2024, down 3.9 percent from its 2023 comparison. It estimated 1,627,900 completions, up 12.4 percent. December's starts were reported at a 1,499,000 seasonally adjusted annual rate. These are different measures: starts describe construction beginning, completions describe units finishing, and an annualized monthly rate is not the number built during that month. The direction of starts and completions could differ because projects move through a pipeline over time.

The practical inference is that an investor could face fresh competing housing even when the pace of new starts was lower than a year earlier. The national figures do not establish where that competition appeared, what rents it achieved, or whether it matched an older property's tenant profile. For a local investigation, list projects within the actual competitive area, their unit types, delivery stages, advertised concessions, and observed leasing progress. Keep the survey's national evidence and the property's local supply evidence in separate columns so one cannot silently stand in for the other.

Worked example: identical principal, three historical rate assumptions

Consider an illustrative $300,000 fully amortizing loan with 360 monthly payments. Applying the three archived benchmark rates purely as arithmetic produces principal-and-interest payments of approximately $2,040 at 7.22 percent, $1,814 at 6.08 percent, and $1,966 at 6.85 percent. The calculation uses monthly interest equal to the annual percentage divided by twelve, with payment equal to principal multiplied by that monthly rate divided by one minus the discount factor over 360 months. The numbers exclude taxes, insurance, mortgage insurance, lender fees, and every operating expense.

The difference between the May and September illustrations is about $226 per month, or $2,716 across twelve payments before rounding. That is meaningful, but it is not additional profit unless the entire property model supports the comparison. A change in points, purchase price, required reserves, or insurance can absorb the apparent benefit. More importantly, the investor cannot assume the most favorable historical rate remained available at closing. A sound reconstructed file shows the dated assumption used before commitment, the financing actually secured, and the effect of any difference on cash reserves and the acquisition decision.

A rental case that remains separate from appreciation

Suppose a hypothetical rental was expected to collect $2,600 monthly. The analyst allows $130 for vacancy and collection loss, $300 for taxes and insurance, $130 for maintenance, $130 for future component replacements, and $208 for management. These invented allowances total $898, leaving $1,702 before debt service and any other property-specific costs. Against the illustrative $1,966 principal-and-interest payment above, modeled cash after those allowances is approximately negative $264 per month. A positive national price index does not change that arithmetic or replenish the operating account.

This example is intentionally transparent rather than representative of a particular market. Replace every allowance with verified local information before using the model for a purchase. Taxes may reset, insurance may depend on condition, and a management agreement may include leasing fees beyond a percentage charge. Replacement reserves also require a component schedule rather than an unexplained percentage. The historical lesson is an inference: an investor could correctly observe national appreciation in 2024 and still buy a property whose income failed to carry its financing. Market direction and operating viability are related questions, but they require separate evidence.

Reconstruct an after-repair value without importing later knowledge

A retrospective renovation analysis should begin with a decision date. If the hypothetical offer was made on June 15, 2024, build the initial comparable-sales set from information available by that date. Preserve later sales in a second set used to evaluate the result. Otherwise the exercise rewards the analyst for knowing the future. For each initial comparable, record location, closing date, usable size, bed and bath counts, condition, lot differences, concessions, and the reason it belongs in the comparison. An index is background; a documented comparable sale is property-level evidence.

Then state the adjustments as hypotheses that a reviewer can challenge. If a renovated comparable suggests a higher value, identify which finished features support the adjustment and whether the proposed scope actually delivers them. Use a value range and show proceeds after selling costs, debt payoff, remaining work, and time-related carrying costs. Test a lower exit value and a slower sale together. This does not claim the national market caused a particular outcome. It shows how a 2024 investor could have reduced dependence on a favorable headline by requiring the project to survive a less favorable property-level result.

Builder competition requires an effective-price comparison

An older renovated house and a newly built house can compete for some of the same buyers, but their published asking prices are not a complete comparison. In a historical research file, preserve the dated advertised price alongside whatever documented incentives were actually available, such as closing-cost assistance or an interest-rate buydown. Do not invent a concession because builders elsewhere used one. Inspect differences in location, lot size, warranties, finish quality, association obligations, and the buyer's likely total housing payment. The relevant question is what the customer could choose at the time.

For an original example, imagine a $340,000 resale competing with a $350,000 new house offering a documented $12,000 buyer credit. Subtracting the credit gives a simple $338,000 economic comparison before accounting for financing structure or eligibility; it does not prove the houses are equivalent or that an appraiser will make that adjustment. The example illustrates why a resale underwriting model should include competitive incentives explicitly. A buyer may value immediate monthly relief differently from an equal reduction in price, so both the dollar comparison and the financing terms require review.

Cash timing is the bridge between market research and execution

Build a month-by-month cash schedule for the hypothetical investment rather than placing every receipt and expense into a single total. Record acquisition cash, deposits, scheduled construction draws, loan interest, utilities, insurance, taxes, marketing, and the expected sale or lease date. Then move the exit three months later without moving expenses out of existence. A property can have a positive expected final margin and still exhaust the investor's available cash before completion. That failure would not appear in an annual house-price series or a mortgage-rate chart.

For example, an extra three months of $2,400 carrying costs requires $7,200 of additional liquidity. If the same delay also requires $4,000 of unplanned work, the combined cash need is $11,200. These are invented stress assumptions, not observed average 2024 costs. Their purpose is to identify a funding decision before the project starts: committed cash, a documented contingency source, reduced acquisition exposure, or declining the purchase. A reserve is useful only when it is accessible when bills arrive. Expected appreciation, an unapproved refinance, and a hoped-for buyer do not meet that timing requirement.

Evaluate the decision process before judging the result

A strong historical review asks whether the original decision was reasonable under the evidence then available, not simply whether the property eventually sold for more. Compare the original rent estimate with executed leases, the scope with paid invoices, the schedule with completion dates, and projected financing with the signed terms. Record both favorable and unfavorable differences. A profitable result can conceal weak underwriting if an unexpected market move rescued the deal; a disappointing result can occur after a disciplined decision when risks that were identified actually materialize.

Use a short variance record: assumption, evidence date, expected amount, actual amount, difference, explanation, and the process change the difference suggests. Distinguish a modeling error from a changed fact. Forgetting an insurance expense is a model defect; a verified premium later changing is an assumption update. Reusing a later comparable sale in a supposed original valuation is a historical integrity problem. Keeping these categories separate turns a 2024 case into a repeatable learning instrument, because it identifies a specific practice to retain or improve rather than rewarding confidence after the outcome is already known.

A defensible 2024 research packet

The finished packet should contain a one-page decision memo, a dated source ledger, the property model, the comparable-sales file, financing documents, a scope and reserve schedule, and a decision-date version preserved from later edits. Mark every national observation as context, every property input as verified or provisional, and every future condition as an assumption. Include the original release vintage next to annual sales and construction statistics. Where a later revision matters, show both versions and explain which calculation uses which. The aim is traceability from a published observation to an actual investment decision.

The original interpretation of this evidence is that 2024 cannot be reduced to either a universal buying opportunity or a universal warning. National price growth, a changing mortgage benchmark, and different construction measures described different aspects of the market. The useful investor response was to connect them through a property-specific model that could be challenged. A completed historical review should therefore end with concrete thresholds: maximum acquisition cost, minimum cash reserve, supported rent or exit range, acceptable financing terms, and a reason to stop. Those thresholds make the research actionable without turning history into a promise about the next transaction.

RESEARCH & REFERENCES

Keep the evidence close.

FHFA: 2024 Q4 house-price release

Published February 25, 2025; Q4 2023–Q4 2024 purchase-only HPI and geographic comparisons. Release vintage preserved.