Disney World's crowds are down. The number everyone quotes to prove it doesn't work.
Last year a run of articles reported that Walt Disney World had gotten noticeably emptier. Most of them leaned on the same kind of evidence: third-party wait time trackers showing crowd levels at multi-year lows. Disney, meanwhile, was describing attendance as broadly stable.
Both things looked true at once, which is what made it interesting. A posted wait time measures how long a queue is, not how many people walked through the gate. There are several reasons those two numbers could come apart.
So I scraped eleven years of wait data for all four parks and tried to work out which one it was.
I did not get a clean answer. I got something better: two measurement problems sitting underneath the public conversation, and a fairly precise account of what nobody can currently determine.
The setup
Queue-Times.com publishes historical wait data going back to 2014: per-attraction annual averages, plus a daily crowd calendar with park hours, event flags, and weather. Their API only serves live snapshots, so I wrote a scraper for the published statistics and calendar pages. About 620 pages, cached to disk so the parsing could be re-run without hitting the site again.
That gave me 16,682 park-days and 2,022 attraction-year records covering 2015 through September 2026.
The first chart I built showed exactly what the headlines described. Magic Kingdom's published crowd level peaked at 70% in 2022 and had fallen to 33% by 2026, the lowest in the series.
Then I ran a validity check and the whole thing came apart.
Problem one: the roster keeps changing
Queue-Times computes its crowd level from the attractions it tracks. That roster is not fixed. Magic Kingdom went from 32 tracked attractions in 2022 to 43 in 2025.
This matters more than it sounds. Trackers add headliners first and fill in low-wait attractions later, so an expanding roster pulls the average down without a single guest changing their behavior. Adding the PeopleMover to a basket of headliners drops the mean; it doesn't mean the park got quieter.
I could not tell those apart, so I rebuilt the measure.
The fix is the same one used for chained price indexes. For each pair of adjacent years, compute the change using only the attractions tracked in both years, then chain those year-over-year links into a series. Attractions can enter and leave without ever contaminating a comparison, because every comparison is matched.
I ran a strict fixed basket alongside it as a check. The two methods agreed within a few points at three of the four parks.
The decline held. All four parks peak in 2019 and none have recovered. So roster drift was not the explanation.
But the check turned up something else.
Problem two: the metrics is not comparable across years
Magic Kingdom's mean posted wait was 24.45 minutes in 2015 and 23.95 minutes in 2022. Nearly identical. The published crowd levels for those years were 41 and 70.
That is a 29-point gap on the same underlying waits.
I looked for other explanations and found the opposite. All four parks show a 2022 high in the published index. Not one shows a corresponding rise in actual posted waits. And the ratio between the published level and my computed index is stable through 2021, then spikes in all four parks simultaneously.
Four parks moving together in the same year is not a coincidence. It is a methodology change.
The practical consequence: the 2022 peak that framed so much of the coverage does not exist in the wait data. Waits in 2022 were already three years into a decline. The metric said it was the busiest year on record.
I want to be careful here. Queue-Times gives this data away for free and it is genuinely useful for its intended purpose, which is telling you whether today is a good day to visit. Within a single year it works fine. It just was not built to be compared across an eleven-year span, and that is how it has been used.
Problem three: a fifth of the calendar is a different product
Magic Kingdom now runs roughly 80 separately ticketed nights a year. Halloween parties, Christmas parties, After Hours events. On those nights attendance is capped by design and the park closes early to day guests, so crowd levels run 53 to 69 percent below an average day.
Those nights sit inside every annual average.
So part of what looks like falling demand is the growth of a product line where low crowds are the point. Not all of it. But enough that any analysis ignoring it is overstating the decline at the park with the most events, which happens to be the park with the steepest drop.
There is a wrinkle. Queue-Times only started labeling these nights in 2022, though Disney has run parties for decades. I reconstructed the earlier ones from anomalously short operating hours in party season and validated the rule against the years that are labeled: 0.99 precision, 1.00 recall at Magic Kingdom.
The same rule failed completely at Epcot, whose ticketed events do not close the park early. So I scoped the adjustment to Magic Kingdom and said so, rather than applying a rule I knew was wrong three-quarters of the time.
What is actually happening
With the measure rebuilt, the finding holds. Posted waits have declined at all four parks since 2019.
But "Disney World got quieter" turns out to be four different stories.
Animal Kingdom is a step change: a 40-point drop in a single year, then a plateau it has never left. Magic Kingdom is continuous erosion, falling every single year with no floor yet. Epcot bottomed in 2024 and is recovering. Hollywood Studios fell furthest through 2025, and is the only park that ever exceeded its 2019 level, peaking in 2021 when distancing cut ride throughput and produced longer queues from fewer people.
That last one is the whole argument in miniature. Same posted wait, opposite meaning.
The seasonal and weekly patterns held a surprise too. Monday is the busiest day of the week at all four parks, across nine complete years. Sunday is the quietest at Magic Kingdom. Weekends are not the peak anyone assumes they are, because Disney trips run on arrival and departure cycles rather than local weekend rhythms.
Magic Kingdom Daily/Monthly Patterns
Epcot Daily/Monthly Patterns
Hollywood Studios Daily/Monthly Patterns
Animal Kingdom Daily/Monthly Patterns
The part I could not resolve
Between 2019 and 2024, Magic Kingdom's attendance fell about 15 percent according to TEA's industry estimates. Its wait index fell 34 percent. At Epcot the gap is starker: attendance down under 3 percent, waits down 29.
Waits fell roughly twice as fast as attendance at one park and ten times as fast at another. Attendance decline is real, and it is not the main story.
Four things could produce that gap. New capacity absorbing guests. Novelty demand decaying from an unusually high 2019 base, since Pandora, Toy Story Land and Galaxy's Edge had all opened in the three years prior. Lightning Lane routing guests out of the standby queue. And park hopping, which inflates gate entries while splitting each guest's queue-hours across two parks.
I could test one of them. I split Magic Kingdom's attractions into those with Lightning Lane access and those without, and ran the index separately on each group.
The gap widens steadily after 2022, which is consistent with routing. But I do not think it proves it, for three reasons. Gaps of similar size appear in 2015 and 2017, before Genie+ existed. The 2020 and 2021 control group is too small to use. And most attractions without Lightning Lane post 5 to 11 minute waits, and Disney does not post below 5, so they physically cannot fall the way a headliner can. I tried restricting the control to attractions with room to fall and only two qualified.
Suggestive. Not conclusive. Separating these explanations would need standby versus Lightning Lane boarding counts, hourly throughput, and hopper utilization rates. None of that is public.
The takeaway
Posted wait time is becoming a worse proxy for how busy the parks are, and every mechanism pushing it that way has strengthened over the last five years.
Lightning Lane decouples queue length from attendance. Park hopping decouples gate entries from time spent in any one park. Hard-ticket nights, now a fifth of Magic Kingdom's calendar, suppress annual averages by design. And the published index most often quoted in coverage is not built to be read across years.
So when you see a headline about Disney crowds collapsing, sourced to a wait time tracker, the underlying trend is real. The number being used to demonstrate it is measuring something that has drifted from what everyone assumes it measures.
That is a less satisfying conclusion than "attendance is down X percent." It is also the one the evidence supports.
Data, code and full methodology: GitHub and Kaggle. Source data powered by Queue-Times.com.
Comments