Trade Flow Index

Notes

Lunar New Year slowdowns and how to adjust container stack counts

Rows of stacked shipping containers seen from above in a container yard.

Every year around late January or February, stack counts at Yantian, Ningbo-Zhoushan, Nansha and Qingdao drop hard, and every year someone on a desk reads that drop as a demand collapse instead of what it is: a labor holiday. If you're nowcasting throughput off daily stack and vessel-call data, Lunar New Year is the one calendar event that will wreck your read if you don't adjust for it first.

Why the stacks move the way they do

The mechanic is straightforward once you've watched it a few cycles. Factories across the Pearl River Delta and Yangtze River Delta push a final wave of export cargo out in the two to three weeks before the holiday, so you'll often see stack counts and gate moves spike before CNY, not drop. Then the holiday hits: migrant labor goes home for one to two weeks (sometimes longer depending on the province and the year), factories idle, and new cargo stops showing up at the yard. Carriers know this is coming months in advance and blank a chunk of their Asia-Europe and transpacific strings around the holiday window, so vessel calls thin out at the same time stacks are drawing down rather than filling up.

What you get is a two-phase pattern: a pre-holiday build that can look like strong demand, followed by a sharp fall-off that can look like weak demand, both of which are really just the same labor calendar playing out. An analyst reading the raw series without that context will call a false peak and a false trough in the same six weeks.

The date itself moves, and that's what breaks naive year-over-year comparisons. The holiday falls on a different Gregorian date depending on the lunar calendar, sometimes shifting by nearly a month: it landed January 29 in 2025 and February 17 in 2026. Comparing "week 5 this year to week 5 last year" without adjusting for where the holiday actually fell will misalign the comparison before you've even started.

Golden Week and the rest of the holiday calendar

Golden Week, the first week of October, behaves differently and shouldn't be adjusted the same way. It's domestic-facing: factories and ports slow for about a week, but there's no mass labor migration like CNY, so the drawdown is shorter and shallower, and the pre-holiday export push is less pronounced. Treating Golden Week with the same correction you'd use for CNY will overcorrect and flatten a real signal out of the series.

The broader point is that China isn't the only calendar you need to track if you're running a global stack-count feed. Diwali affects throughput at Indian gateway ports. Ramadan and Eid create a softer, longer slowdown at Gulf and North African ports that doesn't map cleanly to a single week. Each holiday has its own shape, duration and labor-market driver, and a single blanket "holiday adjustment" applied across all of them will get every port except the one it was tuned for.

Building a seasonal baseline that holds

The fix is to stop indexing by Gregorian calendar date and start indexing by days relative to the holiday itself. Pull three to five years of daily stack counts for the port you care about, line each one up so day zero is the first day of Lunar New Year (not January 1, not a fixed week number), and overlay the series. You'll see the pre-holiday build and post-holiday trough land in roughly the same relative position every cycle, even though the absolute dates shift. That overlay becomes your seasonal baseline: a band of expected deviation you can subtract from the current reading before you decide whether a move is holiday noise or an actual change in trade flow.

Anchorage queue data is worth watching alongside the stacks here too. A queue that stays elevated well past the point where stacks have recovered to pre-holiday levels is a different signal than a queue that clears on schedule, and it's one you'd miss if you were only looking at the yard.

This is the kind of adjustment that's easy to describe and tedious to maintain by hand across a dozen ports and three or four holiday calendars, which is part of what the daily feed on our port throughput tracking page is built to carry for you. If you're tired of re-deriving the CNY correction every January, it's worth a look.

← Back to the blog