Sharon Smith
← Field notes

Why averaging Arcadia comps produces a number that describes no house

A 1954 original and a 2023 rebuild trade on the same block. Averaging those two produces a figure that accurately describes neither. Screening outliers is not cherry-picking — it is the analysis.

Valuation · 6 min read · · Sharon Smith

Arcadia is the hardest corridor in the valley to value, and the reason is structural rather than seasonal. On a single block you will find an original 1954 ranch on a flood-irrigated lot, unchanged since the second owner, and directly across the street a 2023 ground-up rebuild on an identical lot. Same street. Same irrigation. Same view of Camelback. A spread of several million dollars.

Every automated valuation model handles this the same way: it averages them. And the number that comes out the other side is a real number that describes no actual house on that street.

The statistics problem, stated plainly

An average is a good summary of a population when that population is one thing with variation around a center. It is a bad summary when the population is actually two different things that happen to share a zip code. In the second case the mean lands in the gap between the two clusters, where no member of either group lives.

This is not an exotic statistical insight. It is the first thing anyone who has handled data professionally checks, and it is the reason I look at the distribution of comparable sales before I look at any summary of them.

Parcel map, the comparable set, and the sales that did not survive screening. The exclusions are the part worth arguing about.

What I do instead

  1. Pull the comparable set wide — same corridor, same irrigation status, twelve to eighteen months, no filtering yet.
  2. Plot price per square foot against year built and against whether the structure is original, substantially renovated, or new construction. The clusters become visible immediately; you do not need sophisticated tools to see them.
  3. Identify which cluster the subject property actually belongs to. This is a judgment call and I will explain the reasoning rather than assert the conclusion.
  4. Exclude the other cluster explicitly, with a written reason for each exclusion. Not deleted quietly — listed, with the reason attached.
  5. Adjust within the remaining cluster for lot size, irrigation, orientation and view, garage, and condition.

The output is a range with a stated basis, plus a list of every comparable I threw out and why. You see the exclusions. That is deliberate. A valuation you cannot audit is an opinion wearing a number, and you should be suspicious of it regardless of who produced it.

Flood-irrigated lots hold a canopy and a lawn that unirrigated lots in the same zip code cannot. It is a real value factor and it carries a real obligation.

Flood irrigation, which is its own factor entirely

Much of Arcadia holds flood irrigation rights, which is why there are green lawns and eighty-year-old citrus in the middle of the Sonoran Desert. Irrigated and non-irrigated lots on adjacent streets are genuinely not comparable, and this is one of the most frequently misunderstood factors in the corridor.

  • Verify the right actually attaches to the parcel. Do not infer it from the neighbor having it.
  • Understand the delivery schedule and the assessment, which are not the same thing as a water bill.
  • Grading matters enormously on an irrigated lot. Flood irrigation puts several inches of water across the property on a schedule, and that water needs to move away from the foundation rather than toward it. This is where the moisture survey earns its cost.

The same discipline on the listing side

Sellers get the identical treatment. If your house is an original and the block has been trading rebuilds, I will show you the two clusters and explain which one buyers will place you in — even when that is not the conversation you were hoping for. Pricing off the wrong cluster produces a listing that sits, and a listing that sits accumulates days on market, which is the one thing in this business that is genuinely difficult to undo.

An average of two unlike things describes neither. Screening outliers is not cherry-picking, as long as you show the reader exactly what you removed and why.

Everything I know, published before you hire me.

That is deliberate. It is how you check my work.

All field notesGet in touch