NovAsia

Sometimes the sample changes, not the market

Nova Asia AI assistant

Let’s explore this page together

A higher average can result from a different mix of listings. Separate that change from repricing before presenting a market-growth headline.

This article reflects the named expert’s practical perspective. See NovAsia’s editorial policy for how material is prepared and reviewed.

Before putting a market-growth headline on NovAsia, I would ask whether the same homes have actually become more expensive. A rising average does not establish that by itself.

Consider a small hypothetical list.

Four records can tell two different stories

In the first observation, two apartments cost 80,000 units of currency each and two cost 120,000 each. The average is 100,000. In the second, the cheaper offers leave the list and two 160,000 offers enter. The original 120,000 apartments retain their prices.

ObservationPrices, thousandsAverage, thousands
First80; 80; 120; 120100
Second120; 120; 160; 160140

The average has risen 40%. The homes appearing in both observations have not changed price. Calling this “housing became 40% more expensive” would claim more than the example establishes.

We also do not know why the first offers disappeared. A sale is one possibility, but withdrawal, relocation of the advert or omission from the next observation could produce the same visible result. Disappearance should not automatically be recorded as a transaction.

The second average is still useful. It describes the current visible mix as more expensive. It simply answers a different question from the movement of comparable prices.

Bring the result back to the buyer's segment

A buyer choosing a particular size and area may be unaffected by new expensive homes of another format. The locations, property types and area definitions included in the sample determine whether its movement helps with that buyer's question.

For repeat observations, two separate accounts help: what happened to matching offers, and what entered or left the sample. Even the matching subset has limits; it may overrepresent homes that remained advertised longer.

A modest comparison with those limits visible is sufficient for this purpose. Giving it the name “index” would add no evidence.

In this example, “the sample now contains more expensive offers” is a defensible headline. Four records do not support the more dramatic conclusion that the entire market has risen.

No single summary statistic tells us why the number moved

The mean is especially sensitive to composition. A period with more large homes, expensive districts or premium projects can produce a higher average even if prices are unchanged within each segment. Before reading movement as appreciation, I want to know whether the two samples contain broadly comparable properties.

The median protects against a handful of very expensive outliers, but it does not solve composition on its own. Remove enough lower-priced homes and the median rises too. Price per square metre helps with differences in size, yet location, quality, floor, view, completion status and tenure can still shift the distribution. A convenient statistic is not automatically an index.

For a buyer, it is often better to separate three questions. What is available now in the segment they care about? What happened to comparable listings observed in both periods? And how did the mix itself change — which formats entered or left? Those answers can move in different directions without contradicting one another.

Suppose lower-priced studios disappear from the visible inventory while the remaining studios keep the same asking prices. The practical result for a new buyer is real: there are fewer cheap options to choose from. But that is not the same claim as saying existing studios appreciated. A headline that merges the two can create a false sense of urgency.

Repeated listings create another distortion. If one unit appears through several intermediaries, it gains extra weight in the sample. A change in duplication between periods can move averages even when the underlying set of homes is stable. Knowing what counts as one observation is therefore part of interpreting the trend.

There are times when the data simply cannot support a strong market statement. In those cases a narrower description is more useful: “the current visible sample contains more high-priced offers” or “prices were broadly unchanged among matched listings”. The language is less dramatic because the evidence is narrower, not because the analysis is weaker.

The test I care about is whether the metric can distinguish the event we are claiming. If the same 40% rise could come from repricing, a different mix of homes or duplicate records, the number alone does not identify the cause. Buyers need to understand what changed in their segment, not merely see an arrow pointing upward.

Choose the right specialist