Live data

The CAI is not an average of pollutant indices. The worst single sub-index becomes the score. That is why the table also names the pollutant that set it — a reading of 80 driven by fine dust and one driven by ozone call for different days.

“No data” is not a fault. The station may be under maintenance, or the reading for that hour has not arrived. We do not fill the gap with a neighbouring station’s value, because that would not be this station’s value. Picking a nearby station yourself is the honest substitute.

As AirKorea states, this feed is unvalidated real-time data. Confirmed figures are republished later after review. Good enough to decide whether to go outside; not good enough to cite in a paper or a complaint.

Where the band boundaries come from is in what the fine-dust grades actually mean.

Start with the misreading that costs people money. A low count in a recent month does not mean the market cooled. Article 3 of the Real Estate Transaction Report Act gives sellers 30 days from the contract date to file, so the current and previous month always look half-empty. In Jongno-gu, Seoul, by contract month: 56 in March and 89 in April, but 33 in June and 28 in July. June and July will fill in.

Second, deals that are already dead stay on screen. Article 3-2 gives another 30 days to file a cancellation. The struck-through lines are those — and some lines not yet struck through will be.

Third, the [registered] tag means ownership has actually transferred. Older contract months show a high share; recent months show very little. Under article 186 of the Civil Act ownership passes on registration, and that takes time.

So never price a home off a single line. Which step makes the deal irreversible is set out, article by article, in the order to check things before buying a home.

The base rate is not what a bank buys or sells at. Banks add a spread when you buy and take one off when you sell. That spread is the fee, and cash is wider than a transfer. When the news says “the won is at 1,380”, this is the number they mean — not the one at the counter.

Rates are published on business days only. On weekends and public holidays the previous business day’s figure stands, so an unchanged number is correct, not stale.

How much actually disappears is in the currency-exchange spread, and total remittance cost can be worked out with the overseas transfer cost calculator.

This is the rate the bank applies, not the rate you will be quoted. Yours moves with the guarantee ratio, your income and the product you qualify for. Read it as the ordering of banks, not as your number.

⚠️ This feed went dark on 3 August and came back on 11 August — but only halfway. The applied rate arrives; the base rate, the spread and the guarantee ratio all arrive as zero. A check we had already built (base + spread = applied) caught it, so we publish the applied rate and nothing about how it is composed. Printing those zeros would read as “a bank with no spread”.

Some banks send no value at all. That too is missing data, not a rate of zero, so the widget names them.

Whether jeonse or monthly rent works out cheaper is in jeonse vs monthly rent.

This panel is the odd one out. The three above are values for this moment; this one pools every loan made in the last three months and updates only when the city republishes the file. Sharing a screen does not make it contemporaneous — which is the whole point of this page.

A policy loan is two numbers, not one. The bank sets a rate; the city pays part of the interest. The city’s own material confirms that this fund carries a variable rate and that the 2026 programme is worth 2.02 trillion won. What each bank actually averaged on loans it really made is in this file and almost nowhere else.

⚠️ “Effective” is our subtraction, not a published field: average loan rate minus average subsidy. The subsidy normally runs for a set period, so it may not hold for the whole term.

⚠️ Read the loan count before the rate. When we measured this, the banks with the lowest and the highest effective rates both had a sample of one. Compare only the banks with enough loans behind them, and the spread turns out to be far narrower than it first looks.

What stands between a small business and this money is set out in the first gate in front of small-business policy loans.

All five read the issuing agency’s own API. If the upstream does not answer, neither can this page. When that happens we do not dress an old value up as a new one — we say we could not load it.

Air quality upstream throttles under load, so a failed call is retried once, quietly. If that also fails we show the last value we held and label it as such.

All five APIs are called from our server only. Keys live in server environment variables and are never sent to the browser. Which region you picked is not recorded.