Median Home Price

A median home price is the midpoint of a set of home sale prices over a stated period and area: order every sale from lowest to highest, and the median is the one in the middle, with as many sales above it as below. It is reported in place of the average because the distribution of home prices is skewed to the right, with a long tail of expensive sales and a floor at zero, so an average sits above the middle of the market and moves sharply when a handful of large sales close. The median is a measure of the typical transaction, not of the typical home and not of any particular home's value.

Median and mean answer different questions. The average, or mean, adds every sale price and divides by the number of sales, so every dollar in every sale affects it. The median ignores the size of the numbers above and below it and reports only the middle position. In a right-skewed distribution the mean sits above the median, and the gap between them is a measure of how much weight sits in the expensive tail. Both are legitimate; the median is the one usually quoted about housing because it is not moved by the top of the market alone.

There is no single "the" median home price. Several organizations publish medians, and the differences are not rounding.

• The U.S. Census Bureau, in its Monthly New Residential Sales release, publishes a median sales price of new houses sold. The survey is "primarily based on a sample of houses selected from building permits," and a sale is counted at contract rather than at closing: as the release puts it, "a 'sale' is defined as a deposit taken or sales agreement signed."
• The National Association of Realtors publishes an existing-home series. In its own description, "the Existing-Home Sales data measures sales and prices of existing single-family homes for the nation overall," with regional breakdowns, and "these figures include condos and co-ops, in addition to single-family homes."
• Repeat-sales indices are a different instrument entirely, discussed below.

Because new construction differs systematically from the existing stock in size, location and age, the new-home median and the existing-home median are not two estimates of one number. They are two numbers.

The mix effect, which is the point of this page. A median describes the homes that sold. If the composition of what sells changes, the median changes, even if every individual house is worth exactly what it was worth last month. The Census Bureau says so plainly in the Explanatory Notes to its own release: "Changes in sales price data reflect changes in the distribution of houses by region, size, etc., as well as changes in the prices of houses with identical characteristics."

That single sentence disposes of most casual use of the statistic. A month in which lending conditions keep first-time buyers out of the market will show a higher median, because the surviving transactions skew expensive. A month in which a large tract of modest new homes completes will show a lower one. In neither case has the market moved in the way a headline will say it has.

Repeat-sales indices are the answer to the mix effect. Rather than comparing whichever homes sold this period against whichever sold last period, a repeat-sales index tracks the same properties across successive transactions. The Federal Housing Finance Agency describes its House Price Index as "a weighted, repeat-sales index, meaning that it measures average price changes in repeat sales or refinancings on the same properties," built by "reviewing repeat mortgage transactions on single-family properties whose mortgages have been purchased or securitized by Fannie Mae or Freddie Mac since January 1975."

That method removes the mix problem and introduces a different limit, which the description states openly: the universe is loans the two enterprises bought or securitized. Cash purchases are outside it, and so is any mortgage too large or otherwise ineligible for them. An index measuring rate of change on a defined subset of properties is not a price level, and it cannot be quoted as one. A median is a dollar figure about this month's transactions; an index is a number about the change in prices for properties that sold more than once.

Three practical limits worth carrying. First, a national median is an average of markets that behave independently, so it describes no particular place. Second, the price series in the Census release is published without seasonal adjustment. The release's seasonally adjusted table marks the median and average sales price columns "X Not applicable" for every month, and labels the price percent changes it does report "Not adjusted change," so a month-over-month comparison carries the season in it. Third, these are sample estimates with stated uncertainty. The Census release attaches a 90 percent confidence interval to each reported price change, with a footnote explaining that where "the 90 percent confidence interval includes zero," there is "insufficient statistical evidence to conclude that the actual change is different from zero." In the release covering July 2026, every reported change in median and average sales price was smaller than its own confidence interval. That is an ordinary result for this series rather than an unusual month, and it is the reason a single month's movement is not news.

Computing a median is arithmetic anyone can check, and doing it twice on the same neighborhood is the fastest way to see the mix effect.

1. List every qualifying sale in the period and area.
2. Sort them by price.
3. Take the middle one. With an even number of sales, average the middle two.

A hypothetical example. In month one, five homes sell in a neighborhood, at $180,000, $210,000, $240,000, $300,000 and $1,100,000. Sorted, the middle value is the third: the median is $240,000. The mean is different: the five prices total $2,030,000, and 2,030,000 ÷ 5 = $406,000. The average describes none of these houses, which is why the median is the one quoted.

Now month two, in the same neighborhood, with no house changing price. The $1,100,000 property does not sell this month, and two more modest homes do. The six sales are $180,000, $190,000, $190,000, $210,000, $240,000 and $300,000. With an even count, the median is the average of the third and fourth: (190,000 + 210,000) ÷ 2 = $200,000. The median has fallen from $240,000 to $200,000, a drop of $40,000, or about 17 percent (40,000 ÷ 240,000 = 0.1667), and not one house is worth a dollar less than it was. Meanwhile the mean has fallen from $406,000 to 1,310,000 ÷ 6 = $218,333, a far larger fall, driven entirely by the absence of the expensive sale.

This is the whole argument for a repeat-sales index, and the whole argument against reading a monthly median as a statement about what homes are worth. All figures are hypothetical and are not any published series' readings.

The median is the right summary statistic for a skewed distribution, and it is routinely asked to do work it cannot do.

What it is good for

• Describing the typical transaction in a market without letting a few expensive sales dominate, which is exactly what a mean fails to do here.
• Comparing markets and periods on a consistent, simple basis that anyone can reproduce from the underlying sales.
• Serving as an input to other measures. Affordability comparisons need a price for the typical transaction, and this is it.
• Timeliness. The transaction medians are published monthly, well before slower measures.

What it cannot do

• It cannot tell you what your house is worth. That is a question about one property and needs an appraisal or a comparative market analysis.
• It moves with the mix of homes sold, so a change in the median is not evidence that prices changed. The Census Bureau's own Explanatory Notes say so.
• It says nothing about the top or bottom of the market, by construction. A stable median is consistent with the entry-level segment moving sharply.
• There is no single national figure to quote. Different publishers cover new homes, existing homes, and repeat sales of the same properties, and the numbers are not interchangeable.
• Monthly changes are frequently smaller than the statistical uncertainty attached to them, which the publishing agency discloses and most coverage omits.