“The typical asking price remains unchanged at 200.000 euros.” For a property report that sounds like an understandable summary. Whether that also means nothing has changed for someone searching is another question. A single central figure can conceal different distributions.
This workshop uses only its own invented asking-price lists. They do not actually describe Austria or Germany. They help to read general market overviews and to sharpen one’s own search question. A personal budget is not treated here as financially verified.
Two lists on the table
The first list contains nine invented prices. In thousand euros they are: 180, 190, 195, 198, 200, 202, 205, 210 and 220. The second list also contains nine invented prices: 80, 100, 120, 150, 200, 250, 280, 300 and 320.
All prices are already ordered by size. Both lists each sum to 1.800 thousand euros. Divided over nine entries the arithmetic mean in both cases is 200 thousand euros. The middle entry is also in both lists the fifth value, i.e. 200 thousand euros.
NIST explains the arithmetic mean as sum divided by count and the median by the position in the ordered data; for an even number the mean of the two middle values is used. [NIST-LAGE] For our odd lists the middle falls exactly on one entry in each case. The definition explains the calculation, not the quality or representativeness of the data.
The search frame shows the difference
The invented person Lea first wants to check which entries lie within an assumed price range up to 210.000 euros. In the first list that is eight out of nine. In the second it is five out of nine. The same mean and the same median did not show this different selection.
That does not mean the first list is overall better. If someone in the example looks at a considerably lower range, they will find no corresponding entries there, while the second list contains several. Which representation is helpful depends on the concrete question.
Lea also cannot select a suitable property from just tallying the budget. Condition, location, areas, running costs and personal prerequisites are intentionally not listed. The workshop only shows how differently the recorded prices can be distributed, even though two central metrics agree.
Making the distribution visible
For its own report the editorial team can mark the individual invented values on a simple number line. Each point stands for a list entry. The first group lies close around the middle, the second spreads over a larger range. No external graphics or actual object photos are needed for this.
Next to it should be the count, the unit and the basis of the data collection. If instead price groups are formed, the editorial team states their exact boundaries. An entry on a boundary must not end up by chance in two groups or in none. The counting rule is fixed before the analysis.
For a public presentation the central figure can continue to be useful. It is complemented by the required distribution information, rather than having to carry an entire market story on its own. A brief explanation of the underlying selection belongs with it.
Do not make unusual values fit
In the second fictitious list the value 320 might stand out. The editorial team does not automatically remove it merely because the mean would then look more pleasant. With real data one would first check whether a recording error, a different unit, a different segment or a genuinely valid entry is involved.
If an error is confirmed, it is corrected in a traceable way. If the entry is valid, its treatment must suit the factual question. A cleaned dataset needs an explained selection rule. The calculation must not be retrofitted to the desired result.
For one’s own asking-price lists the origin of the observation is also relevant. Repeated listings, unknown status changes or differing date points can influence the selection. These points are clarified before a convenient table is turned into an allegedly complete market.
The appropriate follow-up question to a market overview
If you only read a median or an average, you can ask about the count, the segment, the area and the displayed spread. If you are interested in a particular price range, the selection recorded in that range is additionally relevant. The question then is not merely “What is the typical price?”, but “What does the data basis show for my specific search frame and what is it missing?”
A small dataset can only answer this question to a limited extent. Its values are not automatically extrapolated to all offers in a country. If suitable data are missing, the limit remains visible rather than being concealed by a particularly precise decimal place.
Lea therefore does not take a purchase recommendation from the exercise. She takes away a better reading method: look at the central figure, add the distribution and only then check the concrete selection. An unchanged mean can be a truthful statement. It does not follow that everything has remained the same for every person searching.