Blog home

SafeGraph Data Reveals Coronavirus’s Crippling Impact on International Air Travel

March 11, 2020
Ryan Fox Squire

SafeGraph Patterns data shows the troubling impact of panic surrounding the novel coronavirus epidemic on global travel 

At SafeGraph, we believe that having access to clean, accurate, and comprehensive data allows us to understand and solve society’s biggest questions. Understanding the real implications of the coronavirus crisis, especially as it nears pandemic status, is no exception to this rule.

During our most recent SafeGraph Hackathon, two of our data scientists decided to take a closer look at the socio-economic impact of COVID-19 by analyzing SafeGraph Patterns: a dataset of foot-traffic counts and visitor insights for 3.6 million Points-of-Interest (POI) across the U.S. and Canada. More specifically, our data scientists looked at the effect coronavirus had on international airport visits in relation to the growing number of “coronavirus” Google searches. 

This data tells a compelling story of panic as well as the undeniable impact that pandemic-level crises have on shifting of human behavior in real-time.

Before COVID-19: Air travel trends follow consistent pattern 

One thing that’s historically been true about air travel—especially international air travel—is that it’s highly seasonal. More people travel during the warm and sunny summer months when they’ve got vacation days to use up. Just take a look at the foot-traffic data below for San Francisco International Airport (SFO) from the last few years.

Total visits to SFO between 2017-2019 based on data from the SFO Airport Data Portal
SafeGraph Patterns data during the same period

We wanted to know if and when the growing coronavirus epidemic would potentially throw a curveball into these trends. Some airports, like SFO, report detailed data on airport visitors, such as enplanements and deplanements, but that data often gets published with a lag time of several months. In other words, to get access to actual foot-traffic data, we will have to wait until the Spring to see what SFO reports for the Winter travel season. 

This is where SafeGraph Patterns data comes in handy. Our foot-traffic data, which is updated and published on a monthly basis, makes it possible for us to understand and assess impact much more quickly. To do this, we looked at a number of international and domestic airports across the U.S. to quantify changes in normal airport visitation.

As news of COVID-19 spreads, panic sets in and airport visits drop

It’s impossible to know what foot-traffic to airports in 2020 would look like in an alternate non-COVID-19 world. However, we can look at historical SafeGraph Patterns data to make an informed prediction about what foot-traffic to airports should look like in both January and February 2020. That is, if everything in the world were operating under “normal” conditions.

When comparing this prediction to what was actually measured, the differences between predicted and actual reveal themselves quickly and we start to see just how “abnormal” and fragile the world we’re living in today truly is. 

What happened in January was a steep decline in international airport visits compared to what we would expect based on historical data (prediction). Though, oddly enough, we saw a reverse trend—an increase in foot-traffic—to domestic airports. 

The fact that visits to international airports decrease while visits to regional airports increase means that we cannot say that visits to all airports are decreasing. This could indicate “substitution” behaviors on the part of consumers—opting to travel domestically or regionally instead of internationally—or choosing to travel via smaller airports as a means of avoiding international airports (and international travelers) altogether. 

However, since most airport travel takes place at major international airports, the net effect of this analysis is a large decrease in airport visits overall.

There are fewer international airport visits in the U.S. than predicted in January and February 2020, while there was an uptick in domestic (regional) airport visits during the same period.

To understand the true nature of the discrepancy, we wanted to get a better understanding of the historical error (difference) between SafeGraph Patterns data and actual foot-traffic data. You can see below that, beginning in 2020, the difference becomes increasingly negative (negative means fewer visits than expected).

We built a model to forecast airport visits with historical data that accounts for seasonality and time trends. We then compared our predictions to actual measurements to quantify how “unusual” or “different than expected” the data is. The model’s error centers around zero for three years of historical data but becomes increasingly negative in 2020. Negative error means “fewer visits than expected.”

All signs are pointing to a blip in human behavior. The model we developed was trained using several years of historical data, allowing us to quantify how “different” or “unusual” foot-traffic is during any given time period. For the last three years, for example, no period of time has shown substantial abnormalities. We would expect the same would be true for 2020, under normal circumstances, because there would otherwise be no reason for human behavior to change. But then COVID-19 took hold, and all bets were officially off.

January 2020 is wildly different than what we expected. When our model’s predictions and actual measurements are similar, this indicates that people are visiting airports at normal or expected levels (based on historical trends). The greater the difference between the two, the more abnormal and unusual that behavior becomes. Unlike in 2017, 2018, and 2019, actual measurements for early 2020 nose-dive compared to our predictions. This indicates that people are visiting airports at an abnormally low rate.

And while it would still be interesting to end the story here and simply say that international airport foot-traffic is significantly lower than usual, due to the rapid spread of the novel coronavirus putting the world into a stay-at-home inducing panic, it’s not the full story. Or rather, it doesn’t explain the broader social impact of COVID-19’s arrival worldwide.

The risk of information availability and overload

In today’s digital-first world, people have more access to news and information than ever before. (Whether all that information is accurate is an entirely separate question, one that we won’t even attempt to address here.) However, what has become startlingly true is that, as people become more aware or more obsessed with keeping up-to-speed on the latest developments in the fight against the novel coronavirus, they are shifting many of their own behaviors and almost naturally, whether intentional or not, are going into hiding.

Comparing actual international airport foot-traffic data side-by-side against Google search trends results based on the number of search for the keyword “coronavirus” in January and February 2020. The blue and red lines show a striking mirror image, suggesting a potential causal relationship between them.

In fact, as more and more people search for information on COVID-19, we’ve found that there could be a causal link between awareness for all things related to the coronavirus and a decrease in traffic to international airports. In other words, the more people search for “coronavirus,” the less likely they are to take off and travel. Note that both the search terms and the decrease in airports visits have a peak at the end of January—and then return back toward more normal levels. However, starting in March, we start to see Google search trends sky-rocket (unfortunately, SafeGraph data not yet available). If this causal relationship continues to play out, we won’t be surprised if international airport foot-traffic falls off the cliff even more. 

So, what does this tell us? It’s a story about panic. It’s the same story you’re reading on news sources everywhere, just spelled out with real data. As the coronavirus scare permeates all aspects of our life—including the (hopefully temporary) downward spiral of global economies—it’s quickly shifting and redefining day-to-day human behaviors in all sorts of ways. 

Here, we simply looked at how it is affecting international air travel. But we could use the same model to predict the impact on hotels, shopping malls, public spaces, and beyond.

SafeGraph Patterns data is a powerful measure of foot-traffic data and paints a clear picture of how socio-economic trends are shifting dramatically in the face of COVID-19. Overlaying that onto Google search trends tells us two things: 1) that people want to know more about how they can prepare for an inevitable pandemic and 2) that a growing sense of panic-induced fear is causing people to change their behaviors in dramatic, yet also potentially sensible ways.

Browse the latest

Questions? Get in touch with our team of data experts.