Cambodia: Why Property Market Reports Give Different Numbers

The clearest numerical clash is in Phnom Penh condominium supply. Knight Frank’s H2 2025 report states that existing supply reached 63,334 units and links the increase to the handover of two schemes totalling 1,770 units. UPGA’s 2024 outlook uses a chart titled “Total Completed Condominium Unit By Year” and says the sale take-up rate remained at 50%, with more than 42,000 units taken up. If the 50% rate and the “more than 42,000” count use the same denominator, that denominator must exceed 84,000 units. That is already above Knight Frank’s later existing-supply figure, so timing alone cannot reconcile the two.

That does not make either source demonstrably wrong. Knight Frank labels its universe as existing supply and applies its own market segmentation. UPGA places the take-up line on a completed-unit chart, but the public report does not spell out the denominator formula or provide a project-level inclusion list. Without matching project universes, phase treatment and completion rules, the two totals should not be averaged. This is a genuine residual disagreement in the public material: the gap is visible, while the information needed to reconcile it fully is not.

Prices produce a different kind of apparent conflict. The National Bank of Cambodia publishes a Residential Property Price Index with 2020=100. In its March 2026 bulletin, the February 2026 index is 98.7 nationwide and 98.9 for Phnom Penh. Those are index levels, not dollar prices for an apartment. Knight Frank reports US$676 per sq m for new launches in H2 2025, specifically an average selling price advertised by developers over net saleable area. UPGA’s 2024 prime-location chart reports gross sale prices ranging from US$1,795 to US$2,761 per sq m for completed condominiums by segment, and US$1,450 to US$2,660 for projects under construction. The figures look far apart because they answer different questions.

The same caution applies to UPGA’s 50% take-up and Knight Frank’s 3–4% sales rate. UPGA presents its percentage alongside a long-run completed-stock chart and says more than 42,000 units had been taken up. Knight Frank’s chart is explicitly “Sales Rate of Total Units Available and Monitored by Quarter,” and the text refers to the average sales rate of monitored projects. One looks like a cumulative stock measure; the other is a period sales measure within a monitored universe. Subtracting one from the other would create a number with no coherent market meaning.

Official data can differ across versions for another reason: revision. The NBC states that its RPPI methodology was improved and that previously published RPPI data were revised. A historical value in an older official release can therefore differ from the latest series without proving that the older publication was careless. For a current historical comparison, the latest official series should normally be the reference; when auditing an older report, the relevant question is which version was available at the time.

Similar labels can hide different datasets. The rows below use only figures whose meaning can be recovered from original documents; APS numbers are excluded because the official Q1 2026 deck could not be obtained through the public download flow.

“Supply” sounds straightforward until someone has to decide exactly when a unit enters the count. One researcher may include only completed schemes, another may treat all finished phases as stock, a third may count launched units, and a fourth may also show announced pipeline. For a buyer, this is not semantic housekeeping: the chosen stage changes the market size used to benchmark a specific development.

Knight Frank draws a relatively clear boundary around its 63,334 figure. The H2 2025 report calls it existing condominium supply and says Vue Aston and La Vista One River View added 1,770 units after handover. Incoming supply is shown separately, so future stock is not folded into the existing total. UPGA uses a different presentation. Its 2008–2024 chart is titled “Total Completed Condominium Unit By Year,” with a take-up line over the stock bars. The accompanying text says take-up remained at 50% and more than 42,000 units had been taken up, but it does not state the denominator formula separately. If the percentage and the count use the same base, simple arithmetic implies a denominator above 84,000 units. That should not be rewritten as an “official UPGA total”: it is an inferred lower bound, not a labelled stock number in the publication. The useful finding is that UPGA’s base cannot simply be treated as the same 63,334-unit universe reported by Knight Frank.

Timing alone does not solve this one. A later completed-stock estimate would normally be expected to be at least as large as an earlier one. When the later figure is lower, the first suspects are coverage and inclusion rules: districts, project types, phases, stalled schemes and what qualifies as completed or existing. The public versions do not expose enough project-level detail to rebuild the full gap. The defensible conclusion is therefore not that Phnom Penh somehow lost tens of thousands of completed condos, but that the two researchers are counting different universes whose differences are only partly documented.

APS could provide a third benchmark, but the evidence gate matters here. Its official site confirms the Q1 2026 Market Pulse release dated 8 April 2026 and routes the report through a download form; the public page itself does not expose the deck’s numbers and definitions. APS also describes its broader research approach as combining its own transactional database, field surveys and reputable secondary sources. That tells us where the research comes from, but it does not authenticate a specific Q1 2026 condominium total. For that reason, this audit does not lift exact APS figures from a third-party mirror.

For practical comparison, “supply” should travel with four short fields: cut-off date, project stage, geography and phase-inclusion rule. If one of those is missing, a percentage “error” between totals is not defensible. If all four match and the totals still differ, the residual gap becomes a genuine data issue that should remain visible rather than being averaged away.

An index and a price per square metre answer different questions. The National Bank of Cambodia’s RPPI tracks how the price level changes relative to a base period; it does not state what a “typical apartment” costs today. With 2020=100, an index level of 98.7 in February 2026 is a point on the official series. It cannot be read as US$98.70, multiplied by apartment size, or compared arithmetically with a developer’s asking price.

The methodology exists precisely because a housing market does not sell the same mix of homes every month. The NBC’s March 2026 bulletin says it uses a time-dummy hedonic method and links that method to the IMF’s practical RPPI guidance. The IMF highlights the mix problem: a change in the types, sizes or locations of properties transacted can move a simple average even when the price movement of comparable housing is different. That is why index coverage, characteristics, weights and quality adjustment matter.

Knight Frank’s US$676 per sq m is a different measurement altogether. It is the average selling price advertised by developers for new launches in H2 2025, measured over net saleable area. The report also describes a shift toward more affordable Core product. A launch-price average therefore reflects both market pressure and product mix: if a greater share of the period’s launches sits at the low-cost end, the average can fall even though every existing apartment in Phnom Penh has not been repriced by the same amount.

UPGA adds another layer. Its 2024 chart covers prime locations and separates condominiums by market segment and construction stage. Completed units are shown at roughly US$1,795 per sq m for affordable product, US$2,234 for mid-range and US$2,761 for high-end; under-construction figures are US$1,450, US$1,887 and US$2,660 respectively. These are gross sale prices in UPGA’s own market frame. They do not refute Knight Frank’s US$676 figure because the period, location, stage, segment, price definition and area base are all different.

For a buyer, this makes the phrase “prices are falling” incomplete unless the metric is named. A broad index can decline while a particular prime project remains expensive. Cheaper launches can pull down an advertised launch average while completed high-end stock still sits at a much higher level. Conversely, a high per-square-metre figure in a narrow segment does not establish that the whole residential market is rising.

Historical index comparisons need one more check: version. The NBC explicitly states that it improved the RPPI methodology and revised previously published data. If an older market report and a newer official table show different historical values, first establish whether they use the same vintage of the series. If they do not, part of the disagreement is a data-version issue rather than a market move between two dates.

Percentages are easy to mistake for like-for-like measures because they do not carry an obvious physical unit. Yet 50% and 3–4% become comparable only after three questions are answered: what is in the numerator, what is in the denominator, and over what time horizon is the result accumulated? In these reports, those answers differ or are not fully disclosed.

UPGA says the sale take-up rate remained at 50% and that more than 42,000 units were taken up. The statement sits beside a 2008–2024 chart of total completed condominium units with a take-up line. The page layout makes the measure look cumulative rather than quarterly. However, the publication does not separately state a formula such as “units taken up divided by all completed units.” The safe wording is therefore that UPGA reports a 50% rate on its own base; the denominator should not be reverse-engineered beyond what the publication supports.

Knight Frank’s measure is much closer to current sales velocity. Its chart runs quarterly from 2017 to 2025 and is titled “Sales Rate of Total Units Available and Monitored by Quarter.” It plots available units, units sold and an average sales-rate line. The text says the average sales rate of monitored projects fell to about 3–4% in H2 2025. That is a period measure inside a monitored project universe, not the share of every condominium unit ever sold since the market began.

A simple scenario shows why both figures can coexist. A development may have sold half of its total inventory over several years, while selling only a small share of the remaining available stock in a weak quarter. The cumulative ratio barely moves, even as quarterly velocity is low. The reverse can also happen: a small new scheme can sell quickly in one quarter while the historical market-wide stock remains only slowly absorbed.

Coverage adds another boundary. Knight Frank explicitly refers to monitored projects, which is not automatically the whole completed Phnom Penh condominium stock. Sales in schemes outside that monitored set do not have to enter the quarterly rate. UPGA links its percentage to a broader completed-stock chart but does not publish the project list or explicit denominator on the page. So even the word “sales” can sit on different universes.

The useful reading is not that the market is somehow “50% sold” and “selling at 3%” at the same time. One figure describes an accumulated position inside one measurement system; the other describes recent velocity inside another. For a specific development, the relevant evidence would be its own remaining inventory, verified sales over a defined period and the source’s rule for when a unit counts as sold.

One disagreement in the selected material cannot be closed honestly with “different dates.” Knight Frank reports 63,334 units of existing condominium supply in H2 2025, while UPGA’s 2024 statement says take-up was 50% and more than 42,000 units had been taken up. If UPGA’s percentage and count share one denominator, the implied base is above 84,000 units. An older base would then be larger than the later Knight Frank stock figure, which requires a coverage or inclusion explanation.

The public documents do not provide enough detail to finish that reconciliation. UPGA does not state the take-up formula explicitly or publish the complete project list. Knight Frank’s report does not provide the project-level register behind its 63,334 total. Several explanations are plausible—different project stages, treatment of stalled schemes, phase counting, geography or proprietary classification rules—but choosing one as the established cause would go beyond the evidence. The correct status is: the public material does not allow the two bases to be reconciled.

If the purpose is to describe approximate market scale, each figure can be used with its own label and date. If the purpose is to calculate a particular project’s share of total market supply, the result will change materially with the denominator; presenting one answer as definitive would be misleading until the underlying universe is known.

APS presents a different evidence problem. The official page confirms that the Q1 2026 Market Pulse exists, but without the official deck this audit does not transfer exact numbers from a third-party mirror, even where those numbers look plausible. That is not a data conflict; it is a source-verification limit.

The NBC case is the opposite: the reason for historical differences is disclosed. The central bank states that improved RPPI methodology led to revisions of previously published data. Different vintages should therefore be kept with their release dates. An older value may accurately reflect the series available at that time, while the latest official series is the better reference for a current historical comparison. Where the reason for a discrepancy is documented, say what it is; where it is not, keeping two bounded figures is better than inventing precision by averaging them.