Use Korea Real Estate Board statistics to work out how much Seoul apartment prices rose in the year to August 2026, and you get two answers: 11.5% on the sale price index and 15.5% on the transaction price index. Both are official statistics from the same institution, and neither is a wrong number.
After this note you can calculate both yourself, explain why they can differ, and tell which statistic a housing headline is using. It is not a guide to when to buy or sell. The figures were read on 1 November 2026. According to the board's publication calendar (in Korean), the August sale price index came out on 15 September and the August transaction index on 15 October. On the reading date August was the last month with both indices. The calendar lists September for the transaction index and October for the sale price index on 16 November, so by the time this note goes up on 18 November the last month with both may be September. The numbers here are for August, and how to read them does not change.
Same month, same arithmetic, different answers
The arithmetic is simple. Both indices have June 2026 = 100, and the one-year change is (this month's index ÷ the index a year earlier − 1) × 100. Using the indices from the board's R-ONE service, which run to about twelve decimals, the Seoul apartment sale price index went from 91.7398 in August 2025 to 102.3310 in August 2026, a rise of 11.54%, and the transaction price index went from 88.8006 to 102.5629, a rise of 15.50%. To one decimal that is 11.5% and 15.5%; the gap, measured before rounding, is 3.95 percentage points.
ECOS shows the indices to only one decimal. The same calculation on those values gives (102.3 ÷ 91.7 − 1) × 100 = 11.56% and (102.6 ÷ 88.8 − 1) × 100 = 15.54%, so the sale price index reads 11.6% to one decimal. That difference comes from dividing rounded indices. All changes in this note are calculated from the R-ONE indices.
What each one measures
The sale price index comes from the national housing price survey, which is a sample survey. The monthly apartment sample, as revised in April 2025, is 37,619 homes nationwide and 5,513 in Seoul, and each sample home gets a price every month. If it traded that month, the deal is reviewed and used; if not, similar deals and listings in the same complex; failing that, recent cases and listings in nearby complexes. Market-price data and agents' views are also consulted. The index is the geometric mean of the price ratios between two dates (the Jevons method), and regions are combined using housing-stock weights.
The transaction price index has no sample. It draws on reported actual deals and takes the price change in the same home sold at least twice. Apartments in the same complex, size, building and floor group are assumed to be the same home, and the weights are the number of repeat-sale pairs. Because it must wait for deals to be reported, it comes out about 1.5 months behind. The board lists actual deals and immediate movement as its strengths, and the time lag and instability when and where deals are few as its weaknesses. Even the same month differs in timing: the August sale price index is surveyed as of 1 September, while the August transaction index uses deals contracted in August and reported in August or September.
How you average changes the answer
Two fictional complexes show why design changes the number. Prices are in KRW 100 million. Complex A rises 5% from 12 to 12.6; complex B stays at 6. This year there are 4 deals in A and 1 in B; last year there was 1 in A and 4 in B.
Treat the two equally, as a sample might, and (1.05 × 1.00)^(1/2) = 1.0247, a 2.47% rise. Weight by this year's deals and (1.05^4 × 1.00^1)^(1/5) = 1.0398, a 3.98% rise, because deals clustered where prices rose. Compare average deal prices and it goes from (12 + 4 × 6) ÷ 5 = 7.2 to (4 × 12.6 + 6) ÷ 5 = 11.28, up 56.7%. That only reflects which homes sold; no home in this example rose more than 5%. So an average deal price should not be read as a price index.
This is a simplification. The real transaction index solves many repeat-sale pairs by regression, and this note could not confirm whether deals actually clustered in rising complexes in Seoul in 2026. Still, different averaging and weights can clearly split the answer in one market. According to the board's notes, KB's housing survey uses the arithmetic mean of price ratios (the Carli method), and an arithmetic mean is at least as large as a geometric one. The same question applies when a headline uses KB numbers.
Over five years, the path in between was very different
This gap is not just an August matter. From June 2021, where the current ECOS apartment sale price table starts, to August 2026, the overall Seoul rise is almost the same: 15.4% on the transaction index, from 88.9 to 102.6, and 15.6% on the sale price index, from 88.5 to 102.3.
The path in between is quite different. The transaction index fell 19.8% to 71.3 in December 2022, then rose 43.8% to August 2026. The sale price index fell 7.4% to 82.0 in April 2023, then rose 24.9%. The one-year change bottomed at −21.8% in December 2022 on the transaction index and at −11.2% in April 2023 on the sale price index.
In this stretch the transaction index moved more and bottomed four months earlier. Because sample prices also draw on listings, market data and nearby cases, they may come out smoother, while the transaction index is sensitive to which homes traded that month. That is this note's reading, and one cycle cannot be generalised into a rule.
How August 2026 differed by region
Split the same one-year change to August into four groups, and the transaction and sale price indices show 15.5% and 11.5% for Seoul, 9.7% and 6.9% for the Seoul metro area, 5.6% and 3.5% nationwide, and 0.6% and 0.4% outside the metro area. This August the transaction index is ahead in all four.
That is not always so: in Seoul alone, over the 51 months since June 2022 it was ahead in 37 and behind in 14.
How to read a housing price number
First, identify the statistic: the board's sale price index, its transaction index, a KB index or an average deal price. Second, check the region, the housing type and the comparison period; changes from the previous month, the end of last year and the same month last year cannot be compared with one another. Third, check the date. The sale price index comes out around the 15th of the next month, with a separate weekly survey, while the transaction index comes about 1.5 months later, so two numbers in one day's news may have different latest months. When this note was written on 1 November, the sale price index ran to September (Seoul 103.1) and the transaction index to August.
Fourth, calculate both over the same period side by side. Rather than picking one as the right answer, treat a wide gap as a cue to check where the deals took place. The data are in the board's R-ONE (in Korean) and the real estate price index section of the Bank of Korea's ECOS, and the methods are in the board's notes on the national housing price survey and the transaction price index (both in Korean).
Add your perspective.
Share a question, another approach, or something you have tried.
Checking sign-in…
Loading comments…