Median, mean or index: how to read any house price statistic

Three different objects get published under the heading “house prices”, and they are not competing estimates of one quantity. A mean, a median and an index answer different questions, fail in different ways, and cannot be compared with each other. Most confusion about property statistics is a reader comparing two of the three.

What does a mean house price tell you?

Add every price, divide by the count. It uses all the information, which is its strength, and it is moved by every pound of every sale, which is its weakness. One exceptional transaction in a small area shifts it and nothing marks that it has happened.

A mean is the right statistic when you want a total divided by a count: tax receipts per sale, lending per transaction, the size of a market. It is the wrong one for “what does a home cost here”, because nobody bought at the mean.

What does a median tell you instead?

The middle sale: half above, half below. It barely notices the exceptional transaction, which is exactly why it is the headline figure across this network. It is also the statistic a reader intuitively wants when they ask what something costs.

Its weaknesses are two. It discards information, so it needs a reasonable sample before it stabilises. And it describes what sold, so if the composition of what sold changes, the median moves without any individual home changing price. That second one is the deep problem and it has its own article: why an average can fall while every house gets dearer.

What is a house price index measuring?

An index does not state a price at all. It states a change against a base period, set to 100. Its whole purpose is to hold constant the thing a median cannot: the mix of what is being sold.

There are two main families and the difference matters. A hedonic index models price as a function of a property’s attributes and reports what the price of a constant bundle of attributes would have done; the official UK index works this way. A repeat-sales index uses only properties that sold more than once, so the property itself is the control; what that method fixes and what it cannot see is a separate piece.

A third choice sits underneath both and is made far less consciously: whether the figure is in cash or in today’s money. Nearly every property figure you read is nominal, which is fine over a quarter and misleading over a decade.

An index’s weakness is that it answers only the change question. Told that an index is 144, you know the market has risen 44% since the base year and you do not know what a house costs.

Read down the last column and the division of labour is obvious. An index cannot price a house and a median cannot date a turning point, so a publisher that offers only one of them has decided which half of the questions it will answer. Most offer only the median, which is the right choice for a level and the wrong one for a trend.

Which of the three should you use?

What settles most arguments about house price figures?

Levels from a median, changes from an index. If somebody quotes a percentage change computed from two medians, they have measured the market plus whatever happened to the mix between the two dates, and they cannot tell you how much of the answer is which. If somebody quotes an index as though it were a price, they have quoted a number with no units.

Every market on this network publishes a median as its headline and states its sample size beside it, and at least one publishes an index alongside as the check. What each market measures says which is which, and RealScout is the one that publishes both: a median headline, and a repeat-sales index underneath it as the check.