How Outlier Sales Can Distort Suburb Averages

Understand why an average can move quickly

An arithmetic average adds the recorded sale prices and divides by the number of observations. Because every dollar contributes to the total, one transaction far above or below the rest can move the result. The effect is greater when the sample is small. A changed average therefore does not automatically mean that each property in the suburb changed by the same proportion.

The unusual transaction may be perfectly valid. It could represent a property type, land holding, condition or position that is uncommon in that period. Calling it an error or explaining the parties' motives without evidence would be misleading. The first task is to understand what sold and whether the result belongs in the population the report says it measures.

Use the median for a different question

A median orders the recorded results and identifies the middle observation, or the midpoint of the two central observations where the method requires it. The size of the most expensive or least expensive result does not directly pull that middle position in the way it moves an average. This makes the median useful when a distribution is uneven.

The median is not immune to sample change. If the period contains more renovated houses and fewer compact units, the middle sale can shift because a different group transacted. It also says nothing by itself about the features of the seller's home. Average and median answer different summary questions; neither should be described as the price of every typical property.

Check sample size and property mix together

Always read the observation count beside the statistic. An extreme sale has less influence on a large, stable group than on a handful of transactions, while a median based on few records can jump when just one additional result changes the ordering. If the publisher does not disclose the count, record that as a limitation rather than assuming the sample is substantial.

Then divide the sample conceptually by dwelling form, land, condition, age, title context and local position. A suburb period dominated by apartments should not be compared casually with one dominated by detached houses. Apparent price movement may partly reflect a change in what sold. A property-type breakdown can clarify the picture, provided each smaller group still has enough observations to interpret cautiously.

Identify an outlier without inventing a story

There is no universal price threshold that makes a sale an outlier in every dataset. Begin by plotting or ordering the results, then inspect any observation separated from the main group. Verify its source, date, property category and recorded attributes. State the transparent criterion used if a statistical rule is applied, and keep the original result visible rather than quietly deleting it.

Do not infer distress, family transfer, superior views, redevelopment value or another cause unless a reliable public record supports the statement and it is appropriate to use. A withheld price or incomplete attribute remains unknown. For a seller-facing report, neutral wording such as 'unusually high relative to this defined sample' is safer and more informative than speculation about why the parties agreed to it.

Compare the result with and without unusual observations

A useful sensitivity check shows how the chosen summary changes when the unusual observation is included and excluded, while making clear that this is an analytical test rather than permission to erase a genuine sale. Record the sample definition, period, observation count and reason for the comparison. If the conclusion changes materially, the aggregate statistic is fragile and should be presented with that limitation.

Also compare median, average and the range or distribution of observations rather than relying on one centre point. The aim is not to select whichever measure supports the preferred price narrative. It is to understand how much of the reported movement comes from the composition and spread of the dataset. Where the source cannot be reproduced, avoid publishing a precise worked result.

Return from suburb statistics to comparable sales

For an individual appraisal, the more useful question is whether a sale is comparable to the subject property. A high result for a materially different dwelling may explain an aggregate average without supporting the same conclusion for the seller's home. Compare recent settled sales by property type, land or internal area, accommodation, condition, title, parking and immediate location, and explain both similarities and differences.

Use suburb averages and medians as dated background, not as a multiplier or promise. An in-person appraisal can consider the actual property, current competition and the strongest available comparable evidence. If the sample is thin or heavily affected by unusual transactions, say so plainly. Honest uncertainty gives a seller more decision value than a smooth trend line built from mismatched sales.

Use a transparent sensitivity worksheet

A reproducible worksheet can list each observation by a neutral identifier, recorded amount, date, property category and source status. Calculate the published summary from the complete defined sample first. Then flag, rather than delete, any observation being tested as unusual and recalculate the same measure. Record the rule that triggered the test before viewing the preferred outcome. This keeps the exercise analytical and prevents selective removal from becoming a hidden way to manufacture a smoother result.

The worksheet should show the observation count at every stage and retain unknown fields as unknown. If excluding one result changes the headline direction or scale substantially, the appropriate conclusion is that the summary is sensitive to the sample. It is not evidence that the excluded transaction never happened. For seller decisions, follow the sensitivity check with a separate comparable-sales assessment; statistical influence and property comparability are related questions, but they are not the same test.

• Define the full sample before flagging observations.
• State the outlier criterion in advance.
• Show counts and summaries before and after the test.
• Keep appraisal relevance as a separate judgement.