The Price of Staying: How Housing Prices and Borrowing Costs Changed Across DC Condominium and Residential-House Transactions, 2015–2025

Why this question

For many households in the District, the decision to buy — or to stay put rather than move — turns on two numbers: the price of a home and the cost of borrowing to pay for it. Both changed substantially over the past decade, but not in the same way, and not equally for condominiums and houses.

This working paper asks a deliberately narrow question: how did recorded transaction prices for DC condominiums and residential houses change between 2015 and the mid-2020s, and what would a median-priced purchase have implied for monthly borrowing costs at the start and end of the period?

It is the first working paper in The Price of Staying research series. It covers Washington, DC only; Northern Virginia and suburban Maryland remain part of the series' planned scope. The analysis is exploratory. Its purpose is to establish a careful, documented baseline — and to be explicit about what the data can and cannot support — before any stronger claims are made.

Data and workflow

The analysis starts from the District's public property-assessment data and adds three national reference series: consumer prices, mortgage rates and a repeat-sales house price index. Each step is documented so the work can be checked and repeated.

1. 01Transactions. Draw sales recorded in DC's CAMA (Computer Assisted Mass Appraisal) data on Open Data DC (opens in a new tab), keeping sales marked as qualified: 80,985 candidate qualified sales dated 2015–2025.
2. 02Matching. Link each sale by its SSL (square, suffix and lot) identifier to the current CAMA residential and condominium assessment tables.
3. 03Segmentation. Assign matched sales to a condominium or residential-house segment; set aside sales that match neither.
4. 04Inflation adjustment. Express prices in 2024 dollars using annual BLS Consumer Price Index (opens in a new tab) values, through 2024.
5. 05Fixed-basket checks. Recompute the 2015–2024 comparison while holding the 2015 mix of unit sizes — and, separately, of bedroom counts — constant.
6. 06Mortgage scenarios. Model monthly principal and interest on the median-priced purchase in 2015 and 2025 using annual average Freddie Mac PMMS (opens in a new tab) rates.
7. 07Context. Read the results alongside the FHFA House Price Index (opens in a new tab), a repeat-sales benchmark that measures something different from median transaction prices.

About 96% of candidate sales matched a current assessment record: 42,500 to the residential table and 35,268 to the condominium table. The remaining 3,217 sales matched neither table and are excluded from the segment results. Because matching uses today's assessment tables, a property's segment reflects its current classification, not necessarily its classification at the time of sale.

Real prices diverged by property type, 2015–2024

After adjusting for inflation, the median condominium transaction price in 2024 was 15.4% lower than in 2015, measured in 2024 dollars. The median residential-house transaction price was 8.6% higher. In constant dollars, the typical recorded condominium sale became less expensive over the period, while the typical recorded house sale became more expensive.

A change in the median can reflect a change in what sold — smaller units, a different bedroom mix, different buildings or neighborhoods — as much as a change in what comparable homes are worth. The fixed-basket checks address part of that concern by holding the 2015 mix of unit sizes, or of bedroom counts, constant and asking how prices moved within those groups. Comparison cells required at least 30 sales.

Both checks point the same way as the medians. On the size basket, the 2024 real index (2015 = 100) was 82.5 for condominiums and 105.5 for houses; on the bedroom basket, 81.7 and 110.5. Coverage of 2015 sales was 98.8% for the condominium size basket and 100% for the other three.

These checks reduce the role of limited composition differences; they do not remove it. They do not control for neighborhood, building, condition or renovation, and they are not repeat-sales estimates. The results describe how recorded transaction prices changed by segment. They should not be read as appreciation — or depreciation — of identical properties.

Borrowing costs rose faster than prices, 2015–2025

The payment illustration changes both the period and the units. It runs from 2015 to 2025 and is expressed in nominal dollars, because it is meant to show what a buyer would have faced at the time rather than a constant-dollar comparison. It should not be compared directly with the 2015–2024 real figures above.

In nominal terms, the median condominium price was $470,000 in 2025, compared with $443,750 in 2015. The median residential-house price was $990,000, compared with $685,000.

Assuming a 20% down payment and a 30-year fixed-rate loan at the annual average Freddie Mac rate, the modeled monthly principal-and-interest payment on a median-priced condominium rose from $1,664 in 2015 to about $2,400 in 2025. Because the nominal condominium median rose only modestly, most of that increase within the scenario comes from the higher mortgage rate. For a median-priced house, higher prices and higher rates compound: the modeled payment rose from $2,569 to about $5,056, roughly doubling.

What is still being verified

Several definitions that matter for the results are being confirmed with the DC Office of Tax and Revenue (OTR). Until that work is complete, the findings should be treated as provisional.

• The qualified-sale code. The analysis treats the CAMA QUALIFIED value "Q" as a qualified sale. That interpretation has not yet been independently documented by OTR, so the sample is described as candidate qualified sales, not confirmed arm's-length transactions.
• Historical classification. Segments are assigned from the current CAMA assessment snapshot. The open question is whether contemporaneous 2015–2025 classifications can be reconstructed, which would catch later conversions and reclassifications.
• Unmatched sales. The 3,217 candidate sales that match no current residential or condominium record are under review.

If verification changes the sample or definitions, the results will be revised and the changes noted on this page.

What this means

For real-estate practitioners

• Treat condominiums and houses as different markets. Over this period their real transaction prices moved in opposite directions; a single District-wide figure would blur that.
• Price is only half of the affordability picture. Even where real condominium prices fell, the modeled cost of financing a median-priced purchase rose. Separating price movements from rate movements makes conversations with buyers and sellers clearer.
• Read medians carefully. Median sale prices describe what traded, not what an individual home is worth. The fixed-basket checks help, but neighborhood and condition still matter.

For data-science hiring managers

This paper is also a worked example of applied analytics on administrative data:

• Entity resolution with a reported match rate — linking 80,985 transactions to current assessment tables by parcel identifier, with unmatched records counted rather than silently dropped.
• Careful unit handling — CPI-adjusted comparisons with a stated base year, kept separate from nominal scenario modeling.
• Sensitivity analysis — testing a headline result against composition effects, with minimum cell sizes and coverage reported.
• Explicit assumptions and limits — modeled payments labeled as scenarios, causal claims avoided, and an open verification log with the data publisher.

The headline numbers are deliberately modest in what they claim. The durable value is a documented pipeline that can be audited, corrected and extended to Northern Virginia and suburban Maryland.