Realty Census / Research / Agent turnover

What we found

Across 577,684 agent-weeks of roster data, 3,707 agents who had been listed were no longer listed at the next walk: 0.64% of held agents a week. In the same walks 2,572 agents appeared who had not been listed before, 0.60% a week, so the rosters we measured shrank slightly over the period.

The weekly rate was 0.48% in the walk of October 2, 2026, 0.94% in the walk of October 9, 2026 and 0.48% in the walk of October 23, 2026. The middle walk is high mostly because of Keller Williams: an agent there is only counted as gone once their own profile page is gone, so departures from the week before were confirmed a week late. Pooled over the period, Keller Williams lost 0.74% a week.

Weekly drop-off by brokerage

"Agents held" is the brokerage's agents with a published email on its directory at the walks counted, averaged, not the brokerage's total size. 4 smaller brokerages are in every total but not listed (JBGoodwin REALTORS, Christie's International Real Estate Sereno, Winzone Realty, Pinnacle Estate Properties).

Weekly drop-off by state

States with at least 5,000 agent-weeks of data. The state of each agent who dropped off is known; the number of agents held in each state at each walk is not stored, so it is estimated from each brokerage's current mix of states. Treat these as estimates. 348 agents who dropped off had no published state and count nationally only.

Walk by walk

Changed brokerage, or left?

Dropping off a roster is not the same as changing brokerage. It also covers agents who left the business, went inactive, or whom the brokerage removed late, and a brokerage's site change can remove or restore many profiles at once. We can link a departure to a new brokerage only when the agent appears on another roster we read under the same email address: that happened 10 times in these walks. Agents who move usually take a new email at the new brokerage, so this link catches few moves, and we do not match people by name, because a shared name is not proof.

Method

What we read. The public "find an agent" directories of US brokerages, listing every agent with a published email. Each walk re-reads a brokerage's whole directory. An agent is identified by their email address.

How often. A baseline walk on September 25, 2026, then walks on Oct 1 to October 2, 2026, Oct 8 to October 9, 2026, October 23, 2026. Not every brokerage is read in every walk. Each walk's rate is measured from that brokerage's previous counted walk, in days, and turned into a weekly rate.

What counts as dropping off. An agent listed at the previous walk whom the brokerage's directory no longer lists, after every page of it was read without error. Pages that failed to load make an agent "unknown", not gone. At Keller Williams an agent counts as gone only when their own profile page is gone.

What counts as new. An email address on the directory that was not listed before. A new listing is not always a newly licensed agent; it can be an agent who joined from elsewhere or whose profile went public. Berkshire Hathaway HomeServices is left out of the new-listing rate because new agents there are not added until their own profile shows them.

The baseline is not in the rates. Before the September 25, 2026 walk the stored dates were not the dates each directory was read, so that walk sets the starting point and is not counted.

Not included. RE/MAX, Century 21, ERA, Better Homes and Gardens and Sotheby's: not re-read after the September 25 baseline (blocked or not yet re-read); Real Brokerage, Coldwell Banker, Epique, West USA, Realty of America and Vanguard: every walk after the baseline was held back; the Michigan licence register: a licence list, not a brokerage roster.

What is held back

When one walk would remove more than 8% of a brokerage's agents, a broken read or a site redesign is far more likely than a real exodus, so nothing is removed and that brokerage's walk is left out of every figure here. 24 brokerage walks, at 11 brokerages, were held back this way:

Limits

• Four weeks of data from 14 brokerages. It is a measurement of this period, not a forecast, and it is not an annual turnover rate.
• The rosters measured lean toward large networks; independent brokerages are barely represented.
• An agent who drops off and comes back with a new email counts as one departure and one new listing.
• Brokerages remove departed agents at different speeds, so part of the difference between brokerages is how their sites work.
• State figures use an estimated denominator, as explained above.
• No agent names, emails or profile links are published here; every figure is a count.

Questions this page answers

What share of real estate agents leave their brokerage each week?

On the 14 US brokerage rosters we re-read between September 25, 2026 and October 23, 2026, 0.64% of listed agents dropped off per week, about 1 in 156. Dropping off a roster includes agents who changed brokerage, left the business or were removed late, so it is not a count of brokerage changes alone.

How many new agents join a brokerage each week?

0.60% of held agents per week newly appeared on the same rosters (Berkshire Hathaway HomeServices left out, because new agents there are not added until their own profile shows them).

Which brokerages lose agents fastest?

Among brokerages with at least 2,000 agent-weeks of data, Fathom Realty (1.02% a week) and Douglas Elliman (0.97% a week) dropped the largest shares; Brown Harris Stevens the smallest (0.27%). Rosters also differ in how quickly they remove people.

Cite this page

Realty Census (2026). How fast do real estate agents leave their brokerage? Weekly roster re-reads of 14 US brokerages, September 25, 2026 to October 23, 2026. Count dated 2026-10-23. https://realtycensus.com/research/agent-turnover/