Is the fog thinning? The question keeps being put to the one window that barely sees it

Vol. I, No. 10 · San Francisco — Sunday, July 26, 2026

BY SOL BREGMAN, CONSULTING ACTUARY

A man in my trade does not get paid to find a trend. He gets paid to be right about whether there is one — and the honest answer, most mornings, is that there is not, and that saying so plainly is the whole of the job. So when the desk asked me to settle a claim you hear at every dinner table and in a good deal of print — that San Francisco’s fog is thinning, that the summers are turning sunny, that Karl is in retreat — I did what an actuary does with any story about the future. I went and counted the past. The National Weather Service has logged the visibility at San Francisco International every hour for decades; the Iowa mesonet keeps the archive clean and public. Thirty-one years of it, 1995 through 2025, a quarter-million hourly readings. If Karl were leaving, he would leave a track in that file.

He does not. Count every hour the airport could see no farther than a mile — genuine fog, not haze — and it comes to about a third of one percent of all hours: roughly thirty hours a year, and no more. Break the record into decades and you get 0.35 percent in 1996–2005, 0.40 in 2006–2015, 0.26 in the decade just ended. A tidy little dip at the end, and I can see why a person would want to frame it. But I am paid to spoil exactly that framing. The year-to-year swing in this series runs from 0.09 percent to 0.65 — a standard deviation of about 0.16 — which is larger than the entire spread between the decades. A drop you cannot tell apart from ordinary weather is not a trend. It is a Tuesday. The null hypothesis — that nothing is happening — sits right where I like it, undisturbed.

Now the part that matters more than the trend, because it explains why the trend is a mirage. Look at when the airport fogs. Its fog is a winter phenomenon — roughly one percent of December and January hours, and essentially zero from June through August, the very months the whole city thinks of as fog season. That is not an error in the file; it is the physics. Summer’s famous Karl over the bay side is a deck of stratus riding a few hundred feet overhead — a ceiling, not a wall at street level — and it seldom drops the surface visibility at a runway on the lee of the peninsula. When this desk tested the summer version hour by hour last week it measured that ceiling, not visibility, and found the morning overcast just as stubborn as ever. By the airport’s visibility gauge, in other words, Karl scarcely registers, and never in his own season. It is a poor instrument to convict him of retreat.

So I went to where he actually lives. Half Moon Bay, on the open coast, has reported since 2010, and it tells a different city: fog in 5.7 percent of all hours — some 483 a year, sixteen times the airport’s share — and it peaks, correctly, in late summer, near nine percent of September hours. That is the real Karl, on the record. And over its sixteen years he shows no more sign of leaving than at the airport: 2023 was among the foggiest years in the file, 2025 among the lighter, the whole series bouncing inside a band with no floor sliding out from under it. Two coasts of the same county, two lengths of record, one verdict.

The read, and it is a short one. On the bay side across thirty-one years and on the ocean side across sixteen, the claim that the fog is thinning cannot be sustained against the noise; the honest confidence here is high, because the swing so plainly dwarfs any drift. The reason the story survives anyway is that it is nearly always read off the airport — a station on the sunny side of the bay that meets fog a few dozen winter hours a year, in the wrong season, at a sixteenth of the coast’s rate. One caveat, offered because it is the whole of the honesty: surface visibility at two runways is not the entire life of fog, which also drips on the headlands and sits offshore and buries the Sunset while downtown squints in sun. But every one of those omissions would show more fog than my two gauges catch, not less — so they cut against the thinning story, never for it. The finding, then, is that there is no finding. In my line of work that is not a disappointment. It is the best kind of morning.

BY HAZEL WONG, NUMBERS DESK

San Francisco International moved 54.36 million passengers in 2025, against a peak of 57.70 million in 2018 — recovered to 94.2 percent of it, and no further, six years after the fall. That much is the familiar shape of the city’s decade: back, but not all the way. The useful number is underneath. Sort the passengers into domestic and international and the two halves have gone opposite ways. International travel did not merely recover; it set records — 15.91 million in 2025, above the old 2019 high of 15.24 million, and the second record year running, 2024 having already cleared the bar. The hole is entirely on the domestic side: 38.45 million in 2025, still 11.6 percent under 2018 and 8.7 under 2019. International is now a record 29.3 percent of everyone who passes through, up from under 25 in 2018. The read, and the numbers desk will hold to it: the world came back to San Francisco; America hasn’t, quite. If the domestic cabin leans toward the business day-trip, the convention, the Monday-morning flight, and the international one toward leisure and the reopened routes over the Pacific and to Europe, then the turnstiles are saying the visitor returned and the business traveler is the one still missing — the same emptied-weekday downtown this paper keeps finding by other doors. The caveat that keeps this honest: a passenger count is not a trip purpose, so “domestic equals business” is a lean, not a proof. Confidence high on the counts, which are exact and current to May; moderate on the why. And no, the numbers desk will not call the international line a surge. It set two records without one.

BY GWEN OKAFOR, HOUSING DESK

This desk reads the involuntary exits — the notices, the Ellis filings — the way a coroner reads a wound. But there is a second door out of a San Francisco apartment, and since 2015 the city has kept its receipts. Under the tenant-buyout ordinance a landlord who wants to pay a tenant to go must first hand over a disclosure of the tenant’s rights and file it with the Rent Board; the negotiated payment — cash for keys — is logged when it closes. The file now holds 8,417 of those disclosure filings and 3,650 completed agreements carrying a dollar figure, $169.5 million paid out in eleven years. The typical buyout is smaller than the headlines: a median of $35,000, a mean of $46,440 dragged upward by a long tail that runs to a single $469,562 paid to four tenants in Presidio Heights in 2021. The price has climbed — from about $37,000 in the first years to the low fifties by 2022 and 2024, with 2026 so far averaging past $62,000 — while the count holds remarkably steady, mostly 300 to 390 a year and never even dipping in the pandemic. Where the money goes is the tell. By volume the buyouts cluster in the Sunset and the Mission, the city’s densest banks of rent-controlled flats; by price they run highest in the Castro and Upper Market ($59,700 on average), Noe Valley ($58,100), and Russian Hill ($54,200). The read, offered as read: a buyout is worth to a landlord exactly the gap between what a sitting tenant pays under rent control and what the unit fetches empty — vacant at market, or sold as a condo — so the price of a vacancy runs highest precisely where that gap is widest. It is the same pressure this desk clocked yesterday in the year’s eviction notices, working the other door: where the non-payment notice is the exit forced, the buyout is the check written to avoid forcing it. Two honest caveats. A disclosure filing is not a completed buyout — only about 43 percent of them carry a logged amount, the rest fallen through, declined, or unrecorded — and the file preserves the price, never the pressure that set it. High confidence on the counts and the geography; the rent-gap explanation is a read, not a receipt. Data current to July 25.