My favorite part about getting older is that sometimes I wake up at 3:30 AM for no reason and can’t fall back asleep. This happened this weekend and by 5:30 or so I had exhausted my usual phone games (Wordle, Quordle, Blossom, Immaculate, Strands, etc). Reading my book about the extinction of the dinosaurs and rewatching the 9th inning of a thrilling Texas baseball win only woke me up more.
Then I saw an article from Axios about how travel to the US from other countries fell off a cliff in March. As a true nerd, what excited me about the article was that they linked to an official government data source that I wasn’t familiar with.
I went to the CBP’s Average Wait Time website and found a rich dataset that can be extremely useful for describing current trends. The system is kind of clunky in that you have to select individual agencies and terminals, and the online data appears to only go back for 3 years despite the data collection starting in 2008.
But the data was easily exportable and, with a little bit of elbow grease, you can break down US and non-US passengers being processed by CBP each and every day at the nation’s busiest international airports. That’s really useful for showing if, say, some sort of policy had been implemented that was making people from abroad question whether they should come to the U.S.
Having downloaded the data for 8 of the busiest airports in the country I wanted to show how entries are changing. Travel can be very seasonal so I wanted to compare this year to the same time last year. And I rolled it over 30 days to be able to show any recent changes while not getting lost in the noise of daily change.
The result is below:
The last 30 days in this group of major airports has seen a greater than 10 percent reduction in foreign travelers being processed. I sent out an early version of this graph on Bluesky and some attention was paid to it. This immediately made me nervous because there is always the risk of systemic underreporting any time you’re working with very recent data.
Working with the CBP data immediately made me think about similar concerns when that come to mind when working with crime data — especially early in the year.
I’ve talked before about how much I like the Texas Department of Public Safety Uniform Crime Report website. It gives you a ton of options for downloading pretty much everything Texas is reporting to the FBI as soon as agencies report it to the state.
San Antonio Police Department is one agency that does a great job of posting the most recent completed month’s data early in the following month. The February 2025 data for SAPD, for example, was published very early in March 2025. So if I want to see how thefts this year compare to last year I can easily pull up the YTD NIBRS report and there it is!
Some back of the spreadsheet math points to an 18 percent decline in theft offenses in San Antonio through February 2025 compared to the first two months of 2024.
But you have to be careful when using this data.
Running an Ad Hoc Query on thefts in San Antonio this year clearly shows why. This report lets you break down offenses by type and day. There were 121 thefts reported per day in San Antonio between January 1 and February 18, 105 per day from February 19 through 24, and 88, 67, 75, and 18 on February 25, 26, 27, and 28 respectively.
Law enforcement agencies have over a year until they have to submit final 2025 data to the FBI, so these numbers will be adjusted (usually up). The January numbers are prone to adjustment as well though the amount of change tends to diminish substantially after a month or two.
Theft fell 5 percent in San Antonio in 2024 and the early data from January and February points to a decline so far in 2025. But uncertainty with the data suggests cautioned is warranted when guesstimating what that decline may end up being.
Carl Sagan preached that extraordinary claims require extraordinary evidence. Claiming that theft in San Antonio is falling a lot isn’t exactly extraordinary, but the degree of decline shown in the topline figures screams for more evidence to substantiate it. The closer look helps to evaluate and contextualize the likelihood of a decline so far this year allowing for an appropriate level of analytic caution.
Which brings us back to analyzing air travelers coming to the US.
As I said, I’d never worked with this dataset until very recently, so it made me nervous that my first usage pointed to a huge, sudden, very recent decline. One easy way to figure out if the decline was a real change or a data reporting issue is to do the same process but for US travelers being processed by CBP. I assume that US travelers haven’t changed behavior, so if there’s a huge decline there then we’re probably looking at a data reporting issue.
Here is what we get doing this exercise with New York’s JFK airport.
And LAX:
And Miami:
And just for fun I did Orlando which shows a wild change:
All four airports show an identical trend over the last 30 days with a huge drop in non-US passengers going through CBP processing and an increase in US passengers which doesn’t stand out. This strongly suggests that what we’re seeing is not slow data entry — because I’d expect US and non-US passengers to show the same trend if that was the case.
Then, to be sure it isn’t a data reporting issue I went back the next day and downloaded the last week of CBP processing at JFK. The old data showed 260,250 entries between 3/29 and 4/3/2025 while the newly updated data showed 260,250 entries over that span. An exact match gives very strong confidence that we’re seeing a trend rather than a data issue here.
There are quirks in the data that need to be remembered when contextualizing the data.
Travelers heading to the US from Canadian airports tend to go through customs at the Canadian airport so the data we have now may actually be understating the decline. And some posters on Bluesky pointed out that Easter is much later this year which may mean fewer travelers to Orlando for Spring Break, but no such impact exists with Easter 2024 being earlier than Easter 2023, so I’m skeptical that it’s playing much of a role.
The evidence clearly shows that foreigners have stopped coming to the United States as much as they were at this point last year. This newsletter doesn’t delve into politics, and the reason for this change is both pretty obvious and not in need of further discussion here.
From a data perspective, however, it’s an interesting case study in finding an important trend and working to ensure what you’re seeing is real. This is an issue that deserves close attention as time goes on and the CBP’s dataset is a great way to follow it to see if things get better, worse, or stay the same as the year goes on.
