Published Revisions to Cambodia Housing Price Indicators

An official index has a publication history, not just a current value. In the NBC's Financial Stability Review 2024, the national RPPI for August 2024 is 112.8. In the following annual review, the same August is 105.4. Those numbers do not describe two different market months; they are two official published versions of the same reference period.

The release dates matter as much as the month being measured. The NBC archive dates Financial Stability Review 2024 to 30 May 2025 and Financial Stability Review 2025 to 28 March 2026. Economic and Monetary Statistics Bulletin No. 385, published on 15 May 2026, then states explicitly that the RPPI methodology had been improved and that previously published RPPI data were revised. Its revised series shows December 2024 at 105.5, matching the later annual-review table.

That distinction separates a reference month from a release date. A reference month tells you when the underlying market observation belongs; a release date tells you which published version of the statistics you are reading. When a later release changes an earlier month, the property market did not move again in that historical month. The statistical estimate of that month changed. The NBC launched the official monthly RPPI in June 2022 as a tool for monitoring property-market developments, macroeconomic imbalances and financial-sector risks.

The potential size of this effect was documented before the later official revision. In its 2024 technical-assistance report, the IMF compared the then-published national index with an updated calculation for 2020M1–2021M6: the published series showed 16.0% inflation over that span, while the updated method showed 11.0%. That comparison is not a current market-growth figure. It is evidence that a methodological update can materially reshape the measured history even when the underlying historical period has not changed.

The choice of version therefore depends on the question. If you are reconstructing a decision made in mid-2025, the earlier release should not be silently overwritten by data that were published later; it records the information set available at that time. If you are analysing the historical trajectory from today's standpoint, the revised series is the relevant official history. Both versions are useful, but for different purposes.

The table compares the same national Cambodia RPPI months in two official NBC publications: Financial Stability Review 2024, released on 30 May 2025, and Financial Stability Review 2025, released on 28 March 2026. All twelve later values come directly from the FSR 2025 table; Bulletin No. 385 on 15 May 2026 provides the separate revision evidence, explicitly stating that methodological improvements led to revisions of previously published RPPI data and reproducing December 2024 at 105.5.

The revision was not caused by new transactions somehow occurring in old months. In 2024, the IMF reviewed the RPPI compilation process with the NBC. The index then relied on residential-loan records and used a time-dummy hedonic method with an 18-month window, an approach designed to separate price movement from changes in the mix and characteristics of properties. The broad method was retained, but several parts of the compilation needed improvement.

One set of changes concerned which observations belonged in the index and how the data were cleaned. The IMF noted that the earlier sample could include bank loans used to build a property, even though those records did not represent a purchase of an existing residential property, and it recommended revised filters for land size, bedrooms and floors. A second change concerned stratification: the published series used a coarse split between Phnom Penh and other provinces, while a more granular structure could better reflect differences within the capital. A third concerned outliers, because the previous rule could discard a large share of observations in some months. A fourth involved the hedonic model itself, whose specification the IMF recommended simplifying and strengthening.

The IMF also recommended adding loan data that the NBC had collected from microfinance institutions, particularly to improve representation outside Phnom Penh. The current public NBC materials do not provide enough detail to say precisely which recommendations entered the final revised production series or in what sequence. That matters: it would be too strong to attribute every changed monthly value to one filter, one new dataset or one modelling adjustment. The NBC's own wording is broader — the RPPI methodology was improved, and previously published RPPI data were revised.

Across the twelve 2024 observations shown here, the later official table is between 3.6 and 7.4 index points below the earlier one. August and September show the largest gaps in this window, at 7.4 points each. Those are substantial revisions to the measured level of the series, but they are not price falls that occurred on the day the later report was released. The historical estimate changed; the historical transaction month did not.

The nature of the source data helps explain why a methodological change can move an extended run of historical observations. The RPPI is not a census of every residential sale; in the framework reviewed by the IMF it estimated price change from a loan-based sample and adjusted for property characteristics statistically. When inclusion rules, geographic grouping or the treatment of unusual observations change, historical months can be recomputed on a different statistical footing. A sequence of revised points is therefore not evidence that a new market event occurred across those old months.

Revisions can also have more than one statistical cause. The NBC's broader revision policy lists updated data and data sources, wider coverage, and changes in compilation methods as reasons statistics may be revised. Bulletin No. 385 specifically links the RPPI revision to an improved methodology, yet that statement still does not justify assigning every individual monthly difference to one technical change unless the NBC documents that link separately.

A shared RPPI label and the same 2020=100 base do not make two published versions interchangeable. The base simply fixes the index's normalization point. If the sample filters, strata, outlier treatment or model were changed and the history was recalculated, an earlier 112.8 and a later 105.4 belong to different official versions of the same statistical series. Splicing them together creates a line the NBC itself never published.

For an analysis carried out today, the cleaner approach is to stay inside one coherent later series from beginning to end. An archived release answers a different question: what did the official data show, and what information could a market participant have seen, on a particular date? Both uses are legitimate, but they are not the same task. The revised series is the better basis for today's historical comparison; the archived series is the better record of the information set that existed then.

A revision can change more than the vertical level of the chart. In the earlier table, October 2024 is 111.0 and November is 110.3, so the published month-to-month direction is down. In the later table, the same months are 104.2 and 104.3, making the direction slightly positive. That is a genuine sign change in the short move. February and August also shift from flat in the earlier series to lower than the preceding month in the later one.

The calculation rule follows from that example. Do not take January from an archived release and February from a revised release to compute a monthly change; the same applies to year-on-year comparisons when the two endpoints come from different versions. Such a calculation mixes market movement with a change in statistical treatment. If the only record of an old value is an archived report or screenshot while the continuation exists only in a revised release, show the two versions separately and disclose the break instead of implying a seamless time series.

A practical way to preserve comparability is to store the release identity alongside the observation itself: not merely 'August 2024 — 105.4', but 'August 2024 — 105.4, FSR 2025 released 28 March 2026'. That keeps the provenance of a number visible when a later update arrives. It also matters when reviewing an old investment note: a conclusion may have been reasonable given the official series available then even though the current revised history no longer supports the same wording.

For a buyer, the main consequence of a revision is a change in the broad market context. If the historical level is lower after revision, or a short-term move changes direction, an earlier narrative about market strength may no longer be a good description of the official data. That does not reprice a specific apartment. An RPPI remains an aggregate statistical measure, not an automatic valuation formula for an individual unit.

Coverage is therefore central to interpretation. The IMF's 2024 review described the published RPPI as being built from residential-loan data supplied by commercial banks. At that stage, cash purchases, developer-financed transactions and purchases financed by foreign banks were outside the loan-data sample; observations for apartments and condominiums were too sparse for reliable price measurement, so the published index covered houses. That is a material limitation for someone buying a condominium, but it should not be silently carried forward as a statement about the 2026 index.

The NBC's current statistics page lists the Residential Property Price Index file as updated on 18 May 2026, and Bulletin No. 385 documents both the improved methodology and the resulting revision. Those current public materials do not, however, give enough granular detail on property-type and financing-source coverage to establish whether every limitation described by the IMF in 2024 still applies unchanged. The defensible position is narrower: houses-only published coverage is documented for the methodology reviewed in 2024; the detailed composition of the current sample needs a newer NBC methodological disclosure before it can be stated with the same confidence.

For an individual purchase, the RPPI is one layer of context. It can help frame the official price trajectory for the market segment actually represented in the data, but a unit-level decision still depends on comparable properties and listings, the building's condition and management, title, transaction and holding costs, achievable rent, and exit liquidity. A downward revision of the index cannot be translated mechanically into the same percentage reduction in the value of a chosen apartment.

The archived version still has a legitimate use. It answers the historical question of what the official statistics showed at the time and what information a buyer or analyst could have observed then. The consistent current series is more appropriate for analysing history from today's standpoint, while the archived release preserves the information set available at the earlier date. Keeping those purposes separate is more useful than trying to declare one number permanently 'right' and the other permanently 'wrong'.

For risk assessment, that distinction is more useful than searching for one supposedly final number. The revision is a reminder that a broad market indicator has methodological boundaries and can change as the statistical system improves. The further a particular purchase sits from the transactions represented in the source data — for example, a foreign condominium purchase without local bank financing if that segment is still outside the current sample — the more cautiously the national index should be applied. Until the NBC publishes sufficiently detailed current coverage information, that remains a conditional interpretation, not a claim about the composition of the 2026 index.