RESEARCH · 29 SEP 2026
WEATHER AND THE FASHION CALENDAR
By Pau Almar · Published
Eight years of weekly Google searches for 15 garments in 41 fashion cities show that warm weather leaves total interest in clothing almost unchanged but shifts it from coats towards lighter garments. A clearly warmer autumn delays the coat season by about four days, and in warm years the winter season ends up around ten days shorter.
Every autumn, somewhere in a fashion head office, someone looks at the sales report and says the season hasn't started yet. The coats have been in the stores for weeks, bought many months earlier, and the weather is still warm.
I heard that sentence for years at Zara and at Mango, and I never saw anyone measure it properly. So I measured it myself: eight years of weekly Google searches for 15 garments, set against the weather of every one of those weeks in 41 of the world's leading fashion cities.
What the data says
Warm weather does not reduce total interest in clothing, which barely moves. It changes the mix: in a warm autumn, interest in coats and winter clothing falls, while interest in lighter garments holds up for longer.
A clearly warmer-than-normal autumn pushes the start of the coat season back by about four days. A warm spring brings summer interest forward by about eight. In warm years, the winter season ends up around ten days shorter.
The response follows the afternoon maximum, not the night-time minimum. The reaction concentrates below a daily maximum of about 7.5 °C. Whether the afternoon reaches 12 or 18 degrees changes very little.
The effect appears only in markets with a defined winter. It shows up again, with the seasons reversed, in Johannesburg and Sydney. In most tropical cities the effect is not measurable.
Why it matters
Fashion buys by the calendar, months in advance, and demand arrives with the weather. The coat is the most expensive item in the winter wardrobe. If warm autumns keep coming, the interest doesn't disappear. It shifts towards lighter, cheaper garments, and there are fewer weeks to sell outerwear at full price.
After sixteen years inside Zara and Mango, my reading is that this means lower revenue per customer and more markdowns on outerwear until brands change what they buy. That last part is my interpretation. The study measures searches, not sales.
Data and method
Data. Weekly Google Trends search interest for 15 garment concepts, queried as Knowledge Graph entities so the series are comparable across languages, from May 2018 to August 2026 (426 to 429 weeks per market). Weather comes from the ERA5 reanalysis, 2006 to 2026. Of an initial list of 50 fashion cities, 41 were kept; the six mainland Chinese cities, Porto, Hong Kong and Taipei were excluded, each for a measured reason set out in the paper.
Method. Each weekly weather value is expressed as a deviation from what is normal for that week of the year in that city, using a trend line fitted over twenty years, and divided by that week's own year-to-year spread. Each search series is deseasonalised and standardised against its own history. Eleven hypotheses were written down before the data was examined and tested across 328 model specifications. The intensity models are checked with placebo garments that should not respond to weather (shoes and belts), multiple-testing correction and out-of-sample tests; the timing results are checked by removing one market at a time.
Timing. Across the 28 markets with a defined winter, a one-standard-deviation warmer autumn goes with a 4.18-day later onset of winter search interest (t = 2.95; correct sign in 20 of 28 markets), and a warmer spring with an 8.25-day earlier onset of summer search interest (t = −3.21; 24 of 28). The summer result is the more robust of the two. The winter window shortens by 10.67 days per standard deviation, because it starts later; its end does not move significantly.
Substitution. Total garment interest shows no significant response in 26 of 28 temperate markets, while winter and summer interest move strongly in opposite directions.
Threshold. The winter response concentrates below a daily maximum of about 7.5 °C and is close to flat between 7.5 and 20 °C. The daytime maximum outperforms the minimum and mean temperature in every test.
Robustness. The intensity results hold through the pandemic period, repeat with inverted seasons in the southern hemisphere, and get stronger when four hot, near-tropical markets are removed.
Limitations. The study measures search interest, not purchases. The season-start measure responds to the level of search as well as its timing, and markets hit by the same autumn are not fully independent of each other; the paper sets out both as open questions. A defect in how search terms were grouped affected rainwear data in fifteen markets. The paper reports its own failures in a dedicated chapter.
The full study
The complete paper, with its data sources, methods, every table and the chapter on what went wrong, is available to download below.
// QUESTIONS
How much does a warm autumn delay the coat season, according to VGR's weather study?
A clearly warmer-than-normal autumn pushes the start of the coat season back by about four days. Across 28 markets with a defined winter, a one-standard-deviation warmer autumn goes with a 4.18-day later onset of winter search interest.
Does warm weather reduce overall interest in clothing?
No, total interest in clothing barely moves. In a warm autumn, interest in coats and winter clothing falls while interest in lighter garments holds up for longer.
At what temperature does interest in winter clothing start to respond to the weather?
The reaction concentrates below a daily maximum of about 7.5 °C and is close to flat between 7.5 and 20 °C. The afternoon maximum explains the response better than the night-time minimum.