One month of construction data is not a trend

A headline lands in your inbox on a Friday. Apartment construction starts fell 22% in a single month. A sponsor forwards it with a short note: supply is drying up, rents will follow.

It is worth slowing down before agreeing.

The figure comes from the New Residential Construction report. The Census Bureau and HUD publish it every month, and the August edition came out on September 17. It counts starts, meaning the month a building breaks ground. For apartment investors, the line that matters is buildings with five or more units.

Here is what it showed. Starts in that category ran at an annual rate of 344,000 units in August. In July the rate was 444,000. That is a fall of 22.5%.

"Annual rate" needs a word. Census takes one month of activity, adjusts it for the season, and scales it up to a full year. One slow month becomes a slow year on paper. Large apartment projects make the swings worse. A single big building that breaks ground in July instead of August shifts the count on its own.

You can see the noise in the same series. In May, starts dropped to 282,000. In June they jumped to 519,000. Anyone who read May as a collapse was proven wrong a month later. Census prints a margin of error beside each monthly change for this reason.

A steadier way to read it is to average three months at a time. Using the same data, in thousands of units:

• June to August 2026: 519, 444 and 344. Average: about 436.
• March to May 2026: 481, 494 and 282. Average: 419.
• June to August 2025: 430, 453 and 407. Average: 430.

Read that way, the summer pace of starts was about 4% above the spring. It was about 1% above the same months last year. These numbers show no collapse. They show no boom either.

For a property you own today, starts are not the most urgent figure anyway. A building that breaks ground this month will not compete for your tenants until it is finished and leasing. The count that matters sooner is units under construction. Those are the buildings that will open near you. That count stood at 666,000 in August. In April it was 660,000. The pipeline has barely moved.

So a practical habit. When a sponsor cites a supply number, ask three things. Is it one month, or an average of several? Is it starts, completions, or units under construction? Is it national, or specific to the submarket where the property sits? A national figure can move one way while a single neighborhood moves the other.

None of this means supply does not matter. New buildings shape rents, concessions and renewals for years. It means the evidence should be read at the speed it actually moves, and one month is too short to tell you much.

If a single month of data can swing by a fifth, how many months would you want to see before you believe the trend has turned?