Reading Irish house price trends without being misled

Take every sale in a town, work out the median for each month, draw a line through them, and you will have produced something that looks like a price trend and is not one. The register is superb evidence about individual transactions and a poor instrument for measuring change over time, for three reasons that all pull in different directions at once.

The register is evidence, not an index

The distinction is worth stating plainly because so much commentary ignores it. The register tells you what was paid for specific properties. An index tells you what happened to the price of housing, holding constant the fact that different houses sell in different months. Those are separate questions, and the second one cannot be answered by taking the average of the first without a great deal of care.

This is not a criticism of the register. It was created to publish sale prices under the Property Services (Regulation) Act 2011, and it does that well. Measuring inflation in house prices was never its job, which is why the State has a different body doing that job with the same underlying data.

What the official index does differently

The Central Statistics Office publishes the residential property price index, the official measure of change in Irish residential property prices. It is produced monthly, it adjusts for the mix and characteristics of the properties sold rather than simply averaging them, and it is broken out in ways that matter here, including Dublin against the rest of the country and houses against apartments. The CSO also publishes median prices by area, which are useful precisely because they are labelled as medians rather than dressed up as an index.

The practical rule follows from that. Use the register when you want to know what number a specific house achieved. Use the CSO when you want to know whether prices in your area are rising or falling and by how much. Anyone quoting a percentage change derived from raw register medians has skipped a step, and the next three sections describe what falls through the gap.

Mix, and why a median moves when prices do not

Suppose a town sells mainly three bedroom semis, and one month a scheme of large detached houses closes. The median jumps. Nothing has happened to the value of a three bedroom semi. The following month it falls back and someone writes that the market has corrected. This is composition, not price, and at local scale it dominates everything else because the sample sizes are small.

• A single apartment block completing can drag a local median down for a month even in a strong market.
• A phase of large new houses closing together can lift it for the same reason in the opposite direction.
• A quiet month with four sales produces a median that says more about which four houses sold than about the town.
• Seasonal patterns move the mix as well as the volume, because different kinds of property sell at different times of year.

The defence against all of that is to compare like with like across time rather than everything with everything. Hold the property type constant, hold the rough size constant, and watch what that narrower group does from quarter to quarter. You will lose most of your sample and gain a number that means something, which is a trade almost nobody makes and almost everybody should.

The VAT distortion in new build areas

Ireland has a distortion no neighbouring dataset shares. New dwellings are recorded exclusive of value added tax at 13.5 per cent, second hand homes are not, and the two sit in the same column. In an area where new schemes are completing, the share of VAT exclusive rows changes from month to month, so the median moves as the composition of the sample changes even if every individual price is stable.

Sample size deserves its own warning at local level. A town or a suburb may produce only a handful of qualifying sales in a quarter, and a median calculated from five numbers is not a statistic in any meaningful sense. If you would not draw a conclusion from five people in a survey, do not draw one from five houses, however precise the euro figures look when you write them down.

The recency trap

The third problem is the shape of the file itself. Rows are dated to the day the deed was executed and published after the stamp duty return is filed, so the most recent weeks are always incomplete and always filling in. Chart transaction counts by month and the final months will slope downward for purely administrative reasons. Chart medians and the final months are computed from a partial and possibly unrepresentative sample.

The fix is to exclude the most recent period from any trend line, or to mark it as provisional so that nobody reads it as a signal. The mechanics of the delay are covered in how long until a sale appears, and the same caution applies to any chart built from a file with a filing lag.

Reading a local trend sanely

1. Start with the CSO index for the direction and the rate of change, at the highest level of geography that still describes your market.
2. Use register rows to check whether your specific area is behaving like that wider picture, rather than to calculate a rate of your own.
3. Group by quarter rather than by month, so that small samples become less erratic.
4. Split new from second hand, always, and gross up new build figures if you must combine them.
5. Drop the most recent quarter, or label it provisional and refuse to draw a line through it.
6. Compare like with like across time by holding property type and size roughly constant, using the area data as your starting set.

Follow that and you will usually find your local answer is duller than the headlines and more useful. Most areas most of the time are doing something close to what the region is doing, and the interesting cases are the ones where they are not, which is exactly when it is worth understanding why: a new scheme, a transport change, a school, a supply of one particular kind of home drying up.

Sources: the official measure of price change is the residential property price index published by the Central Statistics Office. Individual transaction evidence comes from the Residential Property Price Register, published weekly by the PSRA from stamp duty returns and covering sales since 2010.