Why we use the median and not the average

Every price figure Suelo publishes is a median. None is an average. The difference matters more than it looks.

What each one does

The average adds every price and divides by the count. The median sorts every price and takes the middle one: half the listings sit above it and half below.

On even data the two are almost identical. On housing data they are not.

The penthouse that moves a municipality

Picture a neighbourhood with nine apartments listed between US$90,000 and US$130,000, and one penthouse at US$3,000,000.

The average of those ten is US$397,000. The median is US$115,000. The average describes one property out of ten; the median describes nine.

The Venezuelan housing market has this shape almost everywhere: many listings clustered in a range and a few far above it. An average in Baruta or Libertador would be permanently inflated by a handful of exceptional properties, and it would move each month according to how many of them happened to be listed rather than according to anything happening in the real market.

The other reason: errors

The prices we crawl come from portals, and portals have typos. A US$120,000 apartment published as US$12,000,000 because of one extra zero is an error we have found and fixed more than once.

An error like that moves a small area's average brutally and permanently. It barely touches the median: an absurd value is still just one position in the sorted line, whether it reads 12 million or 12 billion.

What the median does not fix

The median protects against the extreme value, not against the small sample. With five listings, the median is the third price of five, and it only takes two unusual ones for that third price to represent nothing.

So the median is not our only defence. No area appears in a report with fewer than five listings, and no monthly change is published without at least thirty in each of the two months being compared. The median solves the penthouse problem; the threshold solves the empty-area one.